Shift-Symmetric Scalar Gauss-Bonnet Gravity

• Shift-symmetric scalar Gauss-Bonnet gravity is a modified gravity theory featuring a linear scalar coupling to the Gauss-Bonnet invariant, ensuring that constant scalar shifts alter the action only by a topological term.
• The theory uniquely predicts secondary black-hole hair, broken isospectrality in quasinormal modes, and distinct binary merger dynamics compared to general relativity.
• Numerical and perturbative analyses reveal that the model exhibits modified inspiral rates, intricate ringdown behaviors, and strong-field nonlinear effects that challenge conventional GR predictions.

Shift-symmetric scalar Gauss-Bonnet gravity is the four-dimensional Einstein-scalar-Gauss-Bonnet theory in which the scalar couples linearly to the Gauss-Bonnet density, so that constant scalar shifts change the action only by a topological term and leave the local field equations invariant. In one common normalization its action is

$$S(g,\varphi)=\frac{1}{16\pi}\int d^4x\,\sqrt{-g}\left(R-\frac12 \partial_\mu\varphi\,\partial^\mu\varphi+\alpha \varphi R^2_{\rm GB}\right),$$

with

$$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$

Equivalent conventions also appear with a coupling $2\lambda \phi \mathcal G$. In the contemporary literature, “shift-symmetric Einstein-scalar-Gauss-Bonnet gravity” and “shift-symmetric scalar Gauss-Bonnet gravity” typically denote the same theory class. Its distinctive features are direct curvature sourcing of the scalar, black-hole scalar hair tied to topology and horizon data, a restricted black-hole domain of existence, broken general-relativistic isospectrality in the ringdown spectrum, and strong-field binary dynamics that can differ qualitatively from both general relativity and weak-coupling perturbation theory (Khoo et al., 2024, Ballesteros et al., 9 Oct 2025, Corman et al., 24 Nov 2025).

1. Definition and core structure

The defining property of the theory is the linear coupling function $f(\phi)=\phi$, or equivalently $F(\Phi)=F_0+\alpha \Phi$. Because $\int d^4x\,\sqrt{-g}\,\mathcal G$ is a topological invariant in four dimensions, the transformation $\phi\to \phi+c$ changes the action at most by a topological term. This is the precise sense in which the theory is shift-symmetric. In the canonically normalized EFT notation,

$$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$

the strictly shift-symmetric case is the linear choice $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$ (Herrero-Valea, 2021).

Varying the action yields the generalized Einstein equations

$$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$

with

$$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$0

$$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$1

and scalar equation

$$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$2

Thus the scalar is sourced directly by the Gauss-Bonnet density, and its gradients and second derivatives backreact on the metric equations (Khoo et al., 2024).

This linear model is a particular member of the broader Einstein-scalar-Gauss-Bonnet class with coupling $$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$3. Two comparisons are structurally important. First, Einstein-dilaton-Gauss-Bonnet theory uses the exponential coupling $$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$4; the shift-symmetric theory may therefore be regarded as the linearized version of EdGB in the scalar coupling function, so the two theories agree at sufficiently small scalar amplitude or small effective coupling and diverge at larger coupling (Khoo et al., 2024). Second, a generalized conformal scalar theory can also contain a linear $$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$5 term, but it is not shift-symmetric because the scalar sector also contains exponential conformal couplings, a quartic potential, and a fixed Horndeski completion dictated by conformal invariance of the scalar equation rather than by $$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$6 symmetry (Fernandes, 2021).

2. Scalar charge, black-hole hair, and stationary compact objects

A central distinction of the shift-symmetric theory is between black holes and stars. Stationary black holes generically carry a nonvanishing scalar charge, while stationary neutron stars do not. For stationary, asymptotically flat stars, integrating the scalar equation and using the topological nature of $$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$7 yields $$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$8; by contrast, black holes evade that conclusion and develop asymptotic scalar behavior

$$R^2_{\rm GB}=R_{\mu\nu\rho\sigma}R^{\mu\nu\rho\sigma}-4R_{\mu\nu}R^{\mu\nu}+R^2.$$9

This contrast explains why binary black holes and post-collapse remnants are much more sensitive probes of the theory than quasi-stationary neutron-star inspirals (East et al., 2022).

The scalar charge of stationary black holes is not an arbitrary primary charge. In a covariant differential-form framework, the scalar equation can be converted into a scalar-charge 2-form

$2\lambda \phi \mathcal G$0

For generic $2\lambda \phi \mathcal G$1, its non-closedness is controlled by the obstruction 3-form

$2\lambda \phi \mathcal G$2

In the shift-symmetric case $2\lambda \phi \mathcal G$3, one has $2\lambda \phi \mathcal G$4, so $2\lambda \phi \mathcal G$5 and $2\lambda \phi \mathcal G$6. The scalar charge then satisfies a genuine Gauss law and is fixed by boundary data at the horizon and infinity. For static spherically symmetric black holes,

$2\lambda \phi \mathcal G$7

in the linear normalization used there, and the Prabhu-Stein topological relation quoted in that work is

$2\lambda \phi \mathcal G$8

This is why the scalar hair is described as secondary and horizon-controlled rather than as an independent thermodynamic charge (Ballesteros et al., 9 Oct 2025).

The same conclusion appears in the sensitivity formalism for compact binaries. Defining the black-hole sensitivity by

$2\lambda \phi \mathcal G$9

the first law gives

$f(\phi)=\phi$0

For nonrotating black holes in the shift-symmetric theory, the sensitivity depends only on the invariant combination

$f(\phi)=\phi$1

with endpoint values approximately

$f(\phi)=\phi$2

The isolated-hole bound at $f(\phi)=\phi$3 is

$f(\phi)=\phi$4

and for symmetric binaries the inspiral analysis tightens this to

$f(\phi)=\phi$5

A notable implication is that an adiabatic inspiral can in principle drive one or both black holes outside the domain of existence of the isolated regular branch before merger (Julié et al., 2022).

For rotating black holes, the exact solution space is also bounded. Using the stationary, axisymmetric solutions constructed by Delgado et al., the rotating family is characterized by

$f(\phi)=\phi$6

and exists only on a restricted domain in the $f(\phi)=\phi$7-$f(\phi)=\phi$8 plane bounded by the Kerr line, the static shift-symmetric branch, an extremal branch, and a critical branch. The extremal branch can slightly exceed the Kerr bound $f(\phi)=\phi$9, while the maximal allowed $F(\Phi)=F_0+\alpha \Phi$0 decreases as the spin increases (Khoo et al., 2024).

3. Perturbations, quasinormal modes, and ringdown

The ringdown problem in rotating shift-symmetric Einstein-scalar-Gauss-Bonnet gravity has been formulated directly at the level of coupled metric-scalar perturbations on stationary, axisymmetric black-hole backgrounds. Because the rotating background is not spherically symmetric, perturbations do not separate into exact parity sectors as in Schwarzschild. Instead, one organizes the spectrum into axial-led, polar-led, and scalar-led branches according to dominant multipolar content and continuity to the Kerr limit. After gauge fixing and field redefinitions, the linearized system reduces to seven coupled PDEs in two variables, solved spectrally after compactification and imposition of ingoing horizon and outgoing infinity conditions (Khoo et al., 2024).

The first fully non-perturbative computation of the quasinormal-mode spectrum for rapidly rotating black holes in this specific theory shows several generic features. All EsGB modes connect smoothly to the corresponding Kerr or Schwarzschild modes as $F(\Phi)=F_0+\alpha \Phi$1, but increasingly depart from them as $F(\Phi)=F_0+\alpha \Phi$2 grows. General-relativistic isospectrality is broken: axial-led and polar-led gravitational modes no longer coincide because of scalar hair and scalar-metric coupling. Among the branches computed for $F(\Phi)=F_0+\alpha \Phi$3, the $F(\Phi)=F_0+\alpha \Phi$4 polar-led modes have the lowest scaled real frequency for all considered spins and couplings, whereas the $F(\Phi)=F_0+\alpha \Phi$5 scalar-led modes have the highest scaled real frequency. The damping hierarchy is more intricate because the least damped mode can change character as $F(\Phi)=F_0+\alpha \Phi$6 and $F(\Phi)=F_0+\alpha \Phi$7 vary; near maximal coupling the longest-lived mode can be $F(\Phi)=F_0+\alpha \Phi$8 axial-led, polar-led, scalar-led, or even $F(\Phi)=F_0+\alpha \Phi$9-led. Ringdown dominance is therefore not fixed by the Kerr-like gravitational $\int d^4x\,\sqrt{-g}\,\mathcal G$0 sector alone (Khoo et al., 2024).

The comparison with the quadratic-order weak-coupling perturbation theory of Chung and Yunes is one of the sharpest results. At low spin and small $\int d^4x\,\sqrt{-g}\,\mathcal G$1, the perturbative approximation remains accurate over a substantial part of the allowed coupling interval. At $\int d^4x\,\sqrt{-g}\,\mathcal G$2, the real part of the axial mode remains close to the exact result up to about $\int d^4x\,\sqrt{-g}\,\mathcal G$3 of the limiting coupling, while near $\int d^4x\,\sqrt{-g}\,\mathcal G$4 deviations reach about $\int d^4x\,\sqrt{-g}\,\mathcal G$5 in the imaginary part of the axial mode and about $\int d^4x\,\sqrt{-g}\,\mathcal G$6 in the real part of the polar mode. At $\int d^4x\,\sqrt{-g}\,\mathcal G$7, the discrepancy is much larger: the scaled imaginary part of the axial mode deviates from the weak-coupling result by about $\int d^4x\,\sqrt{-g}\,\mathcal G$8 at $\int d^4x\,\sqrt{-g}\,\mathcal G$9, while the scaled real part agrees reasonably well only up to about $\phi\to \phi+c$0, roughly $\phi\to \phi+c$1 of the limiting coupling. The exact dependence on coupling is therefore especially important for rapidly rotating black holes and for damping rates (Khoo et al., 2024).

The same calculation clarifies what is specific to the shift-symmetric coupling when compared with EdGB. At small $\phi\to \phi+c$2 the two spectra are similar, but EdGB has a smaller domain of existence and the mode spectra diverge more clearly as coupling grows, reflecting the nonlinear effect of the exponential dilatonic coupling (Khoo et al., 2024).

4. Numerical formulations, hyperbolicity, and nonlinear strong-field evolution

A substantial part of the modern development of the theory concerns the problem of evolving the full, nonperturbative equations in dynamical spacetimes. East and Ripley introduced a modified generalized harmonic, or modified harmonic, formulation based on auxiliary Lorentzian metrics $\phi\to \phi+c$3 and $\phi\to \phi+c$4. The gauge constraint is written as

$\phi\to \phi+c$5

and the metric equations are modified so that both coordinate propagation and gauge-constraint propagation are hyperbolic with characteristics set by the auxiliary metrics. In numerical work, a common choice is

$\phi\to \phi+c$6

with

$\phi\to \phi+c$7

This formulation allows stable evolution of the exact shift-symmetric EsGB equations in settings where ordinary generalized harmonic gauge exhibits weak-hyperbolicity problems (East et al., 2020).

These methods make it possible to follow dynamical scalar-cloud formation around initially vacuum black holes, head-on collisions of hairy black holes, and quasi-circular binary inspiral and merger. In isolated black-hole evolutions, scalar hair develops dynamically because

$\phi\to \phi+c$8

forces a nontrivial scalar profile whenever $\phi\to \phi+c$9. In head-on mergers the scalar luminosity can become comparable to the gravitational-wave luminosity for

$$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$0

and stable head-on evolutions were carried out up to about

$$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$1

close to the regime where hyperbolicity is expected to fail outside the horizon in spherical symmetry. The exterior domain can remain hyperbolic even when elliptic regions arise inside horizons, which is why excision is essential in the numerical treatment (East et al., 2020).

For quasi-circular nonspinning binary black holes, fully nonlinear evolutions show that leading-order post-Newtonian theory captures the amplitude of the scalar waveform during inspiral reasonably well, whereas the currently available post-Newtonian phasing is not sufficient for the last few orbits before merger. The scalar field remains close to decoupling-limit scaling during inspiral, and the first nonlinear scalar correction appears only at cubic order in the small-coupling expansion. Near merger there is non-negligible nonlinear enhancement in the scalar field, about $$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$2 in the best equal-mass comparison reported there, but the peak gravitational-wave emission is only weakly modified (Corman et al., 2022).

5. Compact binaries, postmerger matter dynamics, and the slow-merger effect

In binary dynamics the scalar sector enters through both dissipative and conservative channels. Post-Newtonian black-hole binary dynamics depends on the sensitivities $$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$3, the local background fields

$$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$4

and the effective coupling

$$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$5

In the usual weak-field picture, scalar dipole radiation enters at $$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$6PN order and tends to accelerate the inspiral relative to general relativity (Julié et al., 2022, Corman et al., 24 Nov 2025).

Full nonlinear simulations of comparable-mass quasi-circular black-hole binaries reveal that this intuition fails near merger. Using two independent formulations—Modified Generalized Harmonic and modified CCZ4—one finds that although scalar dipole radiation accelerates inspiral in the weak-field regime, the late inspiral and merger can be slower than in GR. In the simulated cases, the scalar flux is roughly two orders of magnitude smaller than the gravitational-wave luminosity near late inspiral, and the gravitational-wave power at fixed frequency is similar in EsGB and GR. The key difference is conservative: the EsGB binary has larger $$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$7, so more energy must be emitted to achieve the same frequency increase. In frequency-domain language,

$$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$8

which implies

$$S=\int d^4x\,\sqrt{|g|}\left(-\frac{M_P^2}{2}R+\frac12 \nabla_\mu\phi\,\nabla^\mu\phi+f\!\left(\frac{\phi}{\Lambda}\right)\mathcal G\right),$$9

in late inspiral. The proposed qualitative picture is therefore a crossover from faster-than-GR inspiral at larger separation to slower-than-GR inspiral near merger (Corman et al., 24 Nov 2025).

Binary neutron star mergers probe a different part of the theory. Because stationary neutron stars do not carry scalar charge, the inspiral remains close to GR even at comparatively large coupling. The postmerger phase is more interesting. When a black hole forms, it develops scalar charge on a timescale of about $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$0–$f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$1 ms, long compared with a single ringdown period. In prompt-collapse examples, the $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$2 ringdown amplitude of $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$3 is about $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$4 smaller than GR at $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$5 km and about $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$6 smaller at $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$7 km. In long-lived remnants, remnant oscillations source scalar radiation, and at couplings near the maximum value for which black holes of the same mass exist, the scalar field exhibits significant nonlinear enhancement. If sufficiently large, that enhancement leads to breakdown of the evolution, seemingly due to loss of hyperbolicity of the underlying equations (East et al., 2022).

These results together imply that the strongest observational consequences of the shift-symmetric theory need not be monotonic inspiral dephasing. Depending on the source class, the characteristic signatures can instead include conservative strong-field slowing of black-hole mergers, delayed growth of black-hole scalar charge after collapse, ringdown amplitude shifts, and postmerger scalar activity in high-curvature remnants (Corman et al., 24 Nov 2025, East et al., 2022).

6. Effective-field-theory status, thermodynamics, and nearby deformations

From the EFT perspective, the scalar-Gauss-Bonnet interaction is a cutoff-suppressed operator. Positivity analyses of scalar-Gauss-Bonnet gravity with a general coupling function constrain the even derivatives of $f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$8, and are highly restrictive for nonlinear scalarization functions. The strictly shift-symmetric linear model,

$f(\phi/\Lambda)=\text{const}+\beta \phi/\Lambda$9

is exceptional because

$$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$0

so those derivative-based positivity bounds are trivially satisfied. This does not remove the generic EFT limitations: the interaction still belongs to a nonrenormalizable expansion, loop effects generate higher-dimension counterterms, and the validity of the theory remains tied to its cutoff (Herrero-Valea, 2021).

Thermodynamics has been the subject of a specific controversy. In a four-dimensional scalar-tensor Einstein-Gauss-Bonnet model with shift symmetry, logarithmic corrections to Wald entropy had previously been attributed to the $$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$1 term. A covariant phase-space analysis argues that once one restores exact shift symmetry at the action level by adding the appropriate boundary term, the would-be logarithmic contribution cancels and the entropy reduces to the standard area law,

$$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$2

In that formulation the first law is preserved by a modified Hawking temperature,

$$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$3

rather than by a modified entropy (Liška et al., 2023).

The covariant scalar-charge framework sharpens the same point from a different angle. Shift symmetry is sufficient to make the scalar charge 2-form closed and to produce a horizon-infinity Gauss law, but it is not sufficient to make the generalized Komar charge closed. The Smarr relation can still contain bulk terms through another 3-form, so scalar-charge conservation and bulk-free thermodynamics are distinct statements (Ballesteros et al., 9 Oct 2025).

The theory also admits technically natural shift-symmetric deformations. A two-scale extension that supplements the linear $$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$4 coupling with a cubic Galileon interaction preserves shift symmetry, modifies the black-hole existence bound, allows smaller black holes than ordinary sSGB, and can screen scalar perturbations near the horizon. In that regime the effective Gauss-Bonnet scale for perturbations is increased by kinetic renormalization, which can strongly suppress dipolar emission constraints while leaving the background scalar charge of sufficiently large black holes nearly unchanged (Thaalba et al., 3 Dec 2025).

Finally, not every theory containing a linear $$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$5 term is shift-symmetric, and not every shift-symmetric higher-derivative scalar-gravity theory contains $$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$6 explicitly. The generalized conformal scalar model is non-shift-symmetric despite its linear scalar-GB coupling (Fernandes, 2021), while a separate renormalizability analysis of derivative-only higher-derivative gravity with exact shift symmetry studies a curvature-squared operator basis without including $$G_{\mu\nu}-\frac12 T_{\mu\nu}^{(\varphi)}+\frac12 T_{\mu\nu}^{(GB)}=0,$$7 itself (Muneyuki et al., 2013). This broader context clarifies that the linear scalar-Gauss-Bonnet coupling is a distinctive but not exhaustive realization of scalar shift symmetry in gravity.

In the current literature, shift-symmetric scalar Gauss-Bonnet gravity is therefore best understood as a specific topological scalar-tensor deformation of general relativity whose scalar hair is compulsory for black holes, secondary rather than primary, and increasingly nonperturbative in the strong-field regime. Its phenomenology is mild for stationary stars, substantial for black-hole ringdown and merger, and sensitive to both hyperbolicity and EFT completion as one approaches the edge of the black-hole existence domain (Khoo et al., 2024, East et al., 2022, Corman et al., 24 Nov 2025).