In order to explain the late-time cosmological acceleration, we propose a new pure geometric gravity based on Deser-Woodard theory [Phys. Rev. Lett. 99, 111301 (2007)], in which the expansion of a matter-dominated universe can be accelerating. This new theory behaves as well as general relativity in solar system and gravitational waves, which is an important improvement over Deser-Woodard nonlocal gravity. Especially, for the simplest case, theoretical considerations determine all the parameters that appear in the Lagrangian.
I. INTRODUCTIONWhich theory is the most beautiful one to explain the late-time cosmological acceleration? Dark energy, e.g., quintessence [1] or phantom [2], is a simple model. However, this mysterious matter is undetectable in the local gravitational systems, and thus, its existence lacks direct verification. Modified gravity is also a good candidate (see [3][4][5][6] for reviews). One can couple new fields to the Einstein-Hilbert Lagrangian or add higher-order geometric terms to the action. But current mainstream modified gravities are not good enough to replace Einstein's theory. The problems come from observations and aesthetics. Many simple models, e.g., f (R) = R + αR 2 theory, can be directly ruled out by observations [5]. In order to fit all known observations, one has to construct complex structures for the Lagrangian, e.g., the Hu-Sawicki model [7]. This always makes the theory not so charming, even though it fits the data well. Another aesthetics problem is that, in order to explain the late-time acceleration, classical theories generally need a parameter related to the Hubble constant H 0 . This may be a potential factor that makes theories suffer from the coincidence problem. We believe if a pure geometric gravity does not introduce any H 0 -related parameters into the Lagrangian and can behave well in various gravitational applications, then it is comparable to previous theories.Pure fourth-order gravity [8] seems to be a good choice. However, combined with Eq. (8) and Eq. (27) in [8] and ρ 0 = O(H 2 0 /G) from dimensional analysis, we obtain σ i = O(c/H 0 ), which is inconsistent with the explanation of [8] that σ i is the size of fundamental particles. Deser-Woodard nonlocal gravity [9] is a pure geometric theory and only introduces dimensionless parameters in the Lagrangian. This theory could explain the late-time acceleration, but gives Ψ = Φ in the weak field limit [see metric (4) for the meaning of Φ and Ψ] for the general formalism [10]. The Cassini mission gives Ψ/Φ = 1 + (2.1 ± 2.3) × 10 −5 in the Solar system [11], and recently, Ψ = Φ has been confirmed at the Galaxy scale [12]. We believe a good gravity theory should give Ψ that exactly equals, not close, to Φ [13]. One thing