Causality, Positivity, and UV Consistency

Higher-curvature coefficients are constrained only after the physical question and its assumptions are fixed. Eikonal time delays test causal propagation on a background, dispersion relations test analytic scattering amplitudes, and principal-symbol calculations test whether an initial-value formulation is strongly hyperbolic. These methods overlap, but none can be replaced by the slogan “the coefficient must be positive.” In particular, a sign derived for a gapped heavy-state completion in dimensions above four is not automatically a four-dimensional graviton positivity theorem.

Required background. Gravitational Actions and Boundary Terms supplies the operator and field-redefinition qualifications, while Curvature supplies the background tensors entering the time delay.

Helpful background. Einstein Equations supplies the GR limit and constraint propagation; Formulations of General Relativity supplies characteristic and hyperbolicity tests; Quantum Gravity supplies the separate ultraviolet-completion questions.

Time delay as a causal diagnostic

Consider high-energy scattering from a weak gravitational shock at impact parameter $b$. In the eikonal regime the phase shift is related to the shift of a null coordinate by

$$\delta(s,b)=-p_v\Delta v.$$

Einstein gravity supplies the reference Shapiro delay. A higher-curvature correction to the graviton three-point vertex adds polarization-dependent terms with more inverse powers of $b$. Camanho, Edelstein, Maldacena, and Zhiboedov show that, when this finite local correction is extrapolated into the range where it competes with the Einstein term, some polarization can acquire a time advance. Their impact-parameter derivation is given in Camanho et al. 2016, §§ 2.1–2.4, eqs. (2.12) and (2.17)–(2.19), and the higher-spin resolution is developed in Camanho et al. 2016, §§ 3.1–4.5, pp. 14–35.

The conclusion is conditional. It assumes an eikonal window, weak coupling, impact parameters larger than the microscopic resolution scale, and enough scale separation to discuss the correction before other operators dominate. Within that window, a finite number of low-spin states cannot repair every polarization channel; a tower of higher-spin states must enter near the higher-curvature scale, or the EFT cutoff must be lowered to that scale. This is primarily a cutoff or new-state conclusion, not a universal choice of sign: changing the sign exchanges which polarization is problematic.

Spectral positivity under explicit UV assumptions

A different argument considers Gauss–Bonnet curvature-squared interactions in $D\gt4$ generated by integrating out a healthy heavy sector. With a positive Källén–Lehmann spectral density $\rho(\mu^2)$ and coupling $y$, the coefficient in the normalization of Cheung and Remmen is

$$\lambda =\frac{y^2}{2}\int_0^\infty \frac{d\mu^2}{\mu^2}\rho(\mu^2)\ge0.$$

The source construction, heavy-state decomposition, and resulting sign are in Cheung and Remmen 2017, eqs. (1)–(16). The assumptions include a tree-level weakly coupled completion, a gapped healthy heavy sector, a positive spectral measure, and a regime in which the induced Gauss–Bonnet interaction dominates the other primordial contributions.

This result is strong within that extension class but narrower than a generic forward-limit dispersion relation for four-dimensional gravity. In four dimensions a constant Gauss–Bonnet term is topological, curvature-squared operators can mix with leading equations under field redefinitions, and the massless graviton produces a forward $t$-channel pole. A subtraction or finite-impact-parameter prescription must be stated before importing ordinary gapped-theory positivity reasoning. More specifically, the graviton pole has residue growing as $s^2/t$, so removing it conflicts with the two-subtraction dispersion relation used for the standard forward bound; the resulting forward limit is otherwise dominated by the pole. This obstruction and the assumptions behind proposed pole-subtracted alternatives are analyzed in Alberte et al. 2020, §§ 2.1–2.2, especially equations (2.3)–(2.7).

Results and open questions as of 2026-08-19

The matrix separates formulations, backgrounds, calculations, results, and limitations. It does not rank every row as equally general.

Formulation or process	Assumptions and exact evidence	Result	Scope, limitation, or unresolved issue
Einstein eikonal baseline	Weak shock, large center-of-mass energy, controlled impact parameter; Camanho et al. 2016, § 2.1, eqs. (2.1)–(2.12)	The leading graviton exchange gives the reference phase and Shapiro delay.	It tests the selected weak background and scattering window, not every spacetime or formulation.
Finite higher-derivative graviton vertex	Weak coupling and a window where the local correction competes with Einstein exchange; Camanho et al. 2016, §§ 2.2–4.5, pp. 8–35	Polarization-dependent advances require a nearby higher-spin tower or a lower cutoff.	Reversing one coefficient does not cure every polarization; the conclusion does not identify a unique UV completion.
Heavy-state Gauss–Bonnet completion in $D\gt4$	Positive spectral density, gapped healthy states, tree-level weak coupling; Cheung and Remmen 2017, eqs. (5)–(16)	The induced coefficient obeys $\lambda\ge0$ in the paper’s normalization.	This does not by itself cover four-dimensional topological Gauss–Bonnet, loops, massless exchanges, or arbitrary strongly coupled completions.
Lovelock and Horndeski principal symbol	Generic weak-field backgrounds in the gauges and formulations analyzed; Papallo and Reall 2017, §§ 2–4, pp. 4–19	Real characteristic speeds are insufficient; generic strong hyperbolicity can fail in the studied systems.	A gauge/formulation result is not a coefficient-positivity theorem and does not exclude every special background or reformulation.
Modified harmonic formulation	Weakly coupled Lovelock and Horndeski theories; Kovacs and Reall 2020, §§ 3.2 and 4.2, PDF	A suitable modified harmonic gauge gives a strongly hyperbolic formulation in the demonstrated regime.	This constructive result resolves the apparent conflict with the preceding gauge-specific obstruction; it does not establish well-posedness at strong coupling, global evolution, or a UV completion.
Four-dimensional massless-graviton forward limit	Twice-subtracted dispersion relation, massless $t$-channel exchange, and its $s^2/t$ residue; Alberte et al. 2020, §§ 2.1–2.2	The graviton pole precludes the standard forward positivity bound; simply deleting it is not a generally justified derivation.	Modified bounds require additional high-energy, impact-parameter, compactification, or Regge assumptions. The result does not say that every gravitational EFT violates analyticity or unitarity.

The strongest conservative reading is to treat the eikonal result as a bound on the EFT’s extrapolation and the spectral result as a conditional sign theorem. A more permissive reading can impose $\lambda\ge0$ on theories that demonstrably satisfy the heavy-state assumptions. These competing positions are not in conflict once dimension, background, cutoff, operator basis, and matter coupling are held fixed.

Independent checks and adverse cases

For the spectral formula, take one stable state,

$$\rho(\mu^2)=Z\,\delta(\mu^2-M^2), \qquad Z\gt0.$$

Then

$$\lambda=\frac{y^2Z}{2M^2}\ge0.$$

It is strictly positive when the coupling $y$ is nonzero.

This independently checks the sign and dimensions, but also exposes every assumption: a negative-norm state would reverse the spectral sign, a massless state would invalidate the gap, and an omitted contact term could change the matching.

For the causal argument, a schematic pair of polarization eigenvalues $\Delta v_{\pm}=\Delta v_{\rm GR}\pm a/b^n$ shows why changing the sign of $a$ only exchanges the advanced eigenchannel once the correction dominates. The GR limit $a\to0$ recovers the common positive delay. The adverse case $b$ approaching the cutoff is precisely where the local truncation cannot be trusted, so it supports a cutoff conclusion rather than an extrapolated time machine.

Finally, hyperbolicity is an independent discriminant. Positivity of a Wilson coefficient does not ensure a complete set of characteristic variables on a curved background, while a strongly hyperbolic low-energy formulation does not prove an analytic unitary UV completion.

Interpretation and open questions

The page establishes conditional implications, not mutable phenomenological bounds. Empirical likelihoods and observations belong to Tests of Gravity; candidate microscopic completions belong to Quantum Gravity. A theory must first pass its own background, branch, hyperbolicity, and cutoff checks before either comparison.

The conclusions could change after a theorem supplies a controlled massless-graviton subtraction in the relevant dimension, or an independent nonlinear calculation changes the causal or hyperbolic conclusion, or a proposed UV completion moves the cutoff and demonstrates matching for the same operator basis and matter coupling. A new numerical bound alone does not alter the theory statement.

Common pitfalls

Combine unlike positivity claims. A spectral sign in a gapped $D\gt4$ construction and an eikonal time-delay obstruction answer different questions.

Treat a time advance beyond the cutoff as a prediction. Its appearance can instead diagnose that the truncated local action was extrapolated beyond its validity domain.

Exercises

Integrate a one-particle spectral density and decide which changes to its assumptions can invalidate the sign. Then analyze the two-polarization delay model and determine whether either sign of $a$ protects both eigenchannels when the correction dominates.

Solution

The delta-function integral gives $\lambda=y^2Z/(2M^2)\ge0$ because $Z\gt0$, $M^2\gt0$, and no independent matching term has been added; equality is possible for $y=0$. In the delay model, $a\gt0$ makes the minus channel smaller and $a\lt0$ makes the plus channel smaller; once $\lvert a\rvert/b^n$ exceeds the Einstein delay, one channel is advanced for either sign. Protecting both channels therefore requires new physics before that regime, not merely reversing $a$.

References

• Alberte, Lasma, Claudia de Rham, Sumer Jaitly, and Andrew J. Tolley. 2020. “Positivity Bounds and the Massless Spin-2 Pole.” Physical Review D 102: 125023. DOI; Open PDF.
• Camanho, Xian O., Jose D. Edelstein, Juan Maldacena, and Alexander Zhiboedov. 2016. “Causality Constraints on Corrections to the Graviton Three-Point Coupling.” JHEP 02: 020. DOI.
• Cheung, Clifford, and Grant N. Remmen. 2017. “Positivity of Curvature-Squared Corrections in Gravity.” Physical Review Letters 118: 051601. DOI.
• Kovacs, Aron D., and Harvey S. Reall. 2020. “Well-Posed Formulation of Lovelock and Horndeski Theories.” Physical Review D 101: 124003. DOI; Open PDF.
• Papallo, Giuseppe, and Harvey S. Reall. 2017. “On the Local Well-Posedness of Lovelock and Horndeski Theories.” Physical Review D 96: 044019. DOI.