The gravitational wave event GW170817 together with its electromagnetic counterparts constrains the speed of gravity to be extremely close to that of light. We first show, on the example of an exact Schwarzschild-de Sitter solution of a specific beyond-Horndeski theory, that imposing the strict equality of these speeds in the asymptotic homogeneous Universe suffices to guarantee so even in the vicinity of the black hole, where large curvature and scalar-field gradients are present. We also find that the solution is stable in a range of the model parameters. We finally show that an infinite class of beyond-Horndeski models satisfying the equality of gravity and light speeds still provide an elegant self-tuning: The very large bare cosmological constant entering the Lagrangian is almost perfectly counterbalanced by the energy-momentum tensor of the scalar field, yielding a tiny observable effective cosmological constant.

We consider Galileon models on curved spacetime, as well as the counterterms introduced to maintain the second-order nature of the field equations of these models when both the metric and the scalar are made dynamical. Working in a gauge invariant framework, we first show how all the third-order time derivatives appearing in the field equations -- both metric and scalar -- of a Galileon model or one defined by a given counterterm can be eliminated to leave field equations which contain at most second-order time derivatives of the metric and of the scalar. The same is shown to hold for arbitrary linear combinations of such models, as well as their k-essence-like/Horndeski generalizations. This supports the claim that the number of degrees of freedom in these models is only 3, counting 2 for the graviton and 1 for the scalar. We comment on the arguments given previously in support of this claim. We then prove that this number of degrees of freedom is strictly less that 4 in one particular such model by carrying out a full-fledged Hamiltonian analysis. In contrast to previous results, our analyses do not assume any particular gauge choice of restricted applicability.

We consider a class of nonlocal, pure-metric modified gravity models which were developed to reproduce the Tully-Fisher relation without dark matter and without changing the amount of weak lensing predicted by general relativity. Previous work gave only the weak field limiting form of the field equations specialized to a static and spherically symmetric geometry. Here we derive the full field equations and specialize them to a homogeneous, isotropic and spatially flat geometry. We also discuss the problem of fitting the free function to reproduce the expansion history. Results are derived for models in which the MOND acceleration a_0 ~ 1.2 x 10^{-10} m/s^{2} is a fundamental constant and for the more phenomenologically interesting case in which the MOND acceleration changes with the cosmological expansion rate.

We show that the current bounds on the time variation of the Newton constant G can put severe constraints on many interesting scalar-tensor theories which possess a shift symmetry and a nonminimal matter-scalar coupling. This includes, in particular, Galileon-like models with a Vainshtein screening mechanism. We underline that this mechanism, if efficient to hide the effects of the scalar field at short distance and in the static approximation, can in general not alter the cosmological time evolution of the scalar field. This results in a locally measured time variation of G which is too large when the matter-scalar coupling is of order one.

We demonstrate how to construct purely metric modifications of gravity which agree with general relativity in the weak field regime appropriate to the solar system, but which possess an ultra-weak field regime when the gravitational acceleration becomes comparable to $a_0 \sim 10^{-10} {\rm m/s}^2$. In this ultra-weak field regime, the models reproduce the MOND force without dark matter and also give enough gravitational lensing to be consistent with existing data. Our models are nonlocal and might conceivably derive from quantum corrections to the effective field equations.

This article provides a general study of the Hamiltonian stability and the hyperbolicity of vector field models involving both a general function of the Faraday tensor and its dual, $f(F^2,F\tilde F)$, as well as a Proca potential for the vector field, $V(A^2)$. In particular it is demonstrated that theories involving only $f(F^2)$ do not satisfy the hyperbolicity conditions. It is then shown that in this class of models, the cosmological dynamics always dilutes the vector field. In the case of a nonminimal coupling to gravity, it is established that theories involving $R f(A^2)$ or $Rf(F^2)$ are generically pathologic. To finish, we exhibit a model where the vector field is not diluted during the cosmological evolution, because of a nonminimal vector field-curvature coupling which maintains second-order field equations. The relevance of such models for cosmology is discussed.

We consider the recently introduced "galileon" field in a dynamical spacetime. When the galileon is assumed to be minimally coupled to the metric, we underline that both field equations of the galileon and the metric involve up to third-order derivatives. We show that a unique nonminimal coupling of the galileon to curvature eliminates all higher derivatives in all field equations, hence yielding second-order equations, without any extra propagating degree of freedom. The resulting theory breaks the generalized "Galilean" invariance of the original model.

This talk reviews the constraints imposed by binary-pulsar data on gravity theories, focusing on ``tensor-scalar'' ones which are the best motivated alternatives to general relativity. We recall that binary-pulsar tests are qualitatively different from solar-system experiments, because of nonperturbative strong-field effects which can occur in compact objects like neutron stars, and because one can observe the effect of gravitational radiation damping. Some theories which are strictly indistinguishable from general relativity in the solar system are ruled out by binary-pulsar observations. During the last months, several impressive new experimental data have been published. Today, the most constraining binary pulsar is no longer the celebrated (Hulse-Taylor) PSR B1913+16, but the neutron star-white dwarf system PSR J1141-6545. In particular, in a region of the ``theory space'', solar-system tests were known to give the tightest constraints; PSR J1141-6545 is now almost as powerful. We also comment on the possible scalar-field effects for the detection of gravitational waves with future interferometers. The presence of a scalar partner to the graviton might be detectable with the LISA space experiment, but we already know that it would have a negligible effect for LIGO and VIRGO, so that the general relativistic wave templates can be used securely for these ground interferometers.

Scalar-tensor theories are the best motivated alternatives to general relativity and provide a mathematically consistent framework to test the various observable predictions. They can involve three functions of the scalar field: (i) a potential (as in "quintessence" models), (ii) a matter-scalar coupling function (as in "extended quintessence", where it may also be rewritten as a nonminimal coupling of the scalar field to the scalar curvature), and (iii) a coupling function of the scalar field to the Gauss-Bonnet topological invariant. We recall the main experimental constraints on this class of theories, and underline that solar-system, binary-pulsar, and cosmological observations give qualitatively different tests. We finally show that the combination of these data is necessary to constrain the existence of a scalar-Gauss-Bonnet coupling.

This talk is based on my work in collaboration with B. Boisseau, D. Polarski, and A.A. Starobinsky. The most natural and best-motivated alternatives to general relativity are the so-called "scalar-tensor" theories, in which the gravitational interaction is mediated not only by a (spin-2) graviton, but also by a (spin-0) scalar field. We study quintessence in this general framework, and show that the microscopic Lagrangian of the theory can be unambiguously reconstructed from two observable cosmological functions of the redshift: the luminosity distance and the linear density perturbation of dustlike matter. We also analyze the constraints imposed on the theory by the knowledge of only the first of these functions, as it will probably be available sooner with a good accuracy. [The main interest of this very short paper is to give the equations in the "Brans-Dicke" parametrization.]

This talk reviews the constraints imposed by binary-pulsar data on gravity theories, and notably on "scalar-tensor" theories which are the most natural alternatives to general relativity. Because neutron stars have a strong gravitational binding energy, binary-pulsar tests are qualitatively different from solar-system experiments: They have the capability of probing models which are indistinguishable from general relativity in weak gravitational field conditions. Besides the two most precise binary-pulsar experiments, in the systems B1913+16 and B1534+12, we also present the results of the various "null" tests of general relativity provided by several neutron star-white dwarf binaries, notably those of gravitational radiation damping. [The main interest of this very short paper is its figure, which also takes into account the "strong equivalence principle" tests.]

We consider scalar-tensor theories of gravity in an accelerating universe. The equations for the background evolution and the perturbations are given in full generality for any parametrization of the Lagrangian, and we stress that apparent singularities are sometimes artifacts of a pathological choice of variables. Adopting a phenomenological viewpoint, i.e., from the observations back to the theory, we show that the knowledge of the luminosity distance as a function of redshift up to z ~ (1-2), which is expected in the near future, severely constrains the viable subclasses of scalar-tensor theories. This is due to the requirement of positive energy for both the graviton and the scalar partner. Assuming a particular form for the Hubble diagram, consistent with present experimental data, we reconstruct the microscopic Lagrangian for various scalar-tensor models, and find that the most natural ones are obtained if the universe is (marginally) closed.

The present acceleration of the Universe strongly indicated by recent observational data can be modeled in the scope of a scalar-tensor theory of gravity. We show that it is possible to determine the structure of this theory (the scalar field potential and the functional form of the scalar-gravity coupling) along with the present density of dustlike matter from the following two observable cosmological functions: the luminosity distance and the linear density perturbation in the dustlike matter component as functions of redshift. Explicit results are presented in the first order in the small inverse Brans-Dicke parameter 1/omega.

Binary systems comprising at least one neutron star contain strong gravitational field regions and thereby provide a testing ground for strong-field gravity. Two types of data can be used to test the law of gravity in compact binaries: binary pulsar observations, or forthcoming gravitational-wave observations of inspiralling binaries. We compare the probing power of these two types of observations within a generic two-parameter family of tensor-scalar gravitational theories. Our analysis generalizes previous work (by us) on binary-pulsar tests by using a sample of realistic equations of state for nuclear matter (instead of a polytrope), and goes beyond a previous study (by C.M. Will) of gravitational-wave tests by considering more general tensor-scalar theories than the one-parameter Jordan-Fierz-Brans-Dicke one. Finite-size effects in tensor-scalar gravity are also discussed.

We study the deflection of light (and the redshift, or integrated time delay) caused by the time-dependent gravitational field generated by a localized material source lying close to the line of sight. Our calculation explicitly takes into account the full, near-zone, plus intermediate-zone, plus wave-zone, retarded gravitational field. Contrary to several recent claims in the literature, we find that the deflections due to both the wave-zone 1/r gravitational wave and the intermediate-zone 1/r^2 retarded fields vanish exactly. The leading total time-dependent deflection caused by a localized material source, such as a binary system, is proven to be given by the quasi-static, near-zone quadrupolar piece of the gravitational field, and therefore to fall off as the inverse cube of the impact parameter.

This talk is based on my work in collaboration with Thibault Damour since 1991. Unified theories, like superstrings, predict the existence of scalar partners to the graviton. Such theories of gravity can be very close to general relativity in weak-field conditions (solar-system experiments), but can deviate significantly from it in the strong-field regime (near compact bodies, like neutron stars). Binary pulsars are thus the best tools available for testing these theories. This talk presents the four main binary-pulsar experiments, and discusses the constraints they impose on a generic class of tensor-scalar theories. It is shown notably that they rule out some models which are strictly indistinguishable from general relativity in the solar system. This illustrates the qualitative difference between binary-pulsar and solar-system tests of relativistic gravity.

Some recently discovered nonperturbative strong-field effects in tensor-scalar theories of gravitation are interpreted as a scalar analog of ferromagnetism: "spontaneous scalarization". This phenomenon leads to very significant deviations from general relativity in conditions involving strong gravitational fields, notably binary-pulsar experiments. Contrary to solar-system experiments, these deviations do not necessarily vanish when the weak-field scalar coupling tends to zero. We compute the scalar "form factors" measuring these deviations, and notably a parameter entering the pulsar timing observable gamma through scalar-field-induced variations of the inertia moment of the pulsar. An exploratory investigation of the confrontation between tensor-scalar theories and binary-pulsar experiments shows that nonperturbative scalar field effects are already very tightly constrained by published data on three binary-pulsar systems. We contrast the probing power of pulsar experiments with that of solar-system ones by plotting the regions they exclude in a generic two-dimensional plane of tensor-scalar theories.