FERMILAB-PUB-07-081-A A discriminating probe of gravity at cosmological scales Pengjie Zhang,1, 2 Michele Liguori,3 Rachel Bean,4 and Scott Dodelson5, 6 1Shanghai Astronomical Observatory, Chinese Academy of Science, 80 Nandan Road, Shanghai, China, 200030 2Joint Institute for Galaxy and Cosmology (JOINGC) of SHAO and USTC 3Department of Applied Mathematics and Theoretical Physics, Centre for Mathematical Sciences, University of Cambridge, Wilberforce Road, Cambridge, CB3 0WA, United Kingdom 4Department of Astronomy, Cornell University, Ithaca, NY 14853 5Center for Particle Astrophysics, Fermi National Accelerator Laboratory, Batavia, IL 60510-0500 ∗ 6Department of Astronomy & Astrophysics, The University of Chicago, Chicago, IL 60637-1433 The standard cosmological model is based on general relativity and includes dark matter and dark energy. An important prediction of this model is a fixed relationship between the gravitational potentials responsible for gravitational lensing and the matter overdensity. Alternative theories of gravity often make different predictions for this relationship. We propose a set of measurements which can test the lensing/matter relationship, thereby distinguishing between dark energy/matter models and models in which gravity differs from general relativity. Planned optical, infrared and radio galaxy and lensing surveys will be able to measure EG, an observational quantity whose expectation value is equal to the ratio of the Laplacian of the Newtonian potentials to the peculiar velocity divergence, to percent accuracy. We show that this will easily separate alternatives such as ΛCDM, DGP, TeVeS and f(R) gravity. PACS numbers: Introduction.— Predictions based on general relativ- equation algebraically relates ∇2φ to the fractional over- ity plus the Standard Model of particle physics are at density δ, so lensing is essentially determined by δ along odds with a variety of independent astronomical obser- the line of sight. This is a prediction of the standard, vations on galactic and cosmological scales. This failure GR-based theory that is generally not obeyed by alter- has led to modifications in particle physics. By introduc- nate theories of gravity. ing dark matter and dark energy, cosmologists have been Testing this prediction is non-trivial. Astronomers of- able to account for a wide range of observations, from ten use the galaxy overdensity as a probe of the underly- the overall expansion of the universe to the large scale ing matter overdensity, but the two are not exactly equal. structure of the early and late universe [1]. Alternatively, Here we propose a test of this prediction which is rela- attempts have been made to modify general relativity at tively insensitive to the problem of galaxy bias. The basic galactic [2] or cosmological scales [3, 4]. A fundamental idea is simple: question then arises: Can the two sets of modifications be distinguished from one another? • Extract the matter overdensity at a given redshift by measuring the velocity field. Matter conser- The answer is “No” if only the zero order expansion vation relates velocities to the overdensities. The of the universe is considered. By allowing the dark measurement of the velocity field can be accom- energy equation of state wDE to be a free function, plished by studying the anisotropy of the galaxy the expansion history H(z) produced by any modified power spectrum in redshift space. arXiv:0704.1932v1 [astro-ph] 16 Apr 2007 gravity can be mimicked exactly. Fortunately, struc- ture formation in modified gravities in general differs • Extract the lensing signal at this redshift by cross- [5, 6, 7, 8, 9, 10, 11, 12] from that in general relativ- correlating these galaxies and lensing maps recon- ity. The difference we focus on here is the relationship structed from background galaxies. between gravitational potentials responsible for gravita- tional lensing and the matter overdensity. Lensing is More quantitatively, the galaxy-velocity cross power sensitive to ∇2(φ − ψ) along the line of sight where φ spectrum Pgθ ≡ −hδg(k)θ(−k)i can be inferred from and ψ are the two potentials in the perturbed Friedman- redshift distortions in a galaxy distribution. Here, θ ≡ Robertson-Walker metric: ds2 = (1 + 2ψ)dt2 − a2(1 + ∇ · v/H(z) and v is the comoving peculiar velocity. In 2φ)dx2 and a is the scale factor. In standard general the linear regime, matter conservation relates θ to δ by ˙ relativity (GR), in the absence of anisotropic stresses, θ = −δ/H = −βδ, where β ≡ d ln D/d ln a and D is φ = −ψ, so lensing is sensitive to ∇2φ. The Poisson the linear density growth factor. So, Pgθ = βPgδ , sat- isfying the first goal above. Cross correlating the same galaxies with lensing maps constructed from galaxies at higher redshifts, P∇2(φ−ψ)g can be measured. The ra- ∗Electronic address: [email protected],[email protected],[email protected],[email protected] of these two cross-spectra therefore is a direct probe 2 2 of ∇ (φ − ψ)/(βδ). It does not depend on galaxy bias Here, W is the lensing kernel. For a flat universe, χ, χs or on the initial matter fluctuations, at least in the lin- are the comoving angular diameter distance to the lens ear regime. Modifications in gravity will in general leave and source, respectively. Eq. 3 is a pure geometric result signatures in either β and/or the Poisson equation. and can be applied to any modified gravity models where Galaxy-Velocity Cross-correlation.— A galaxy’s photons follow null geodesics. peculiar motion shifts its apparent radial position from A standard method to recover the redshift information s xz to xz = xz +vz/H(z) in redshift space, where vz is the is by the lensing-galaxy cross correlation. For galaxies in comoving radial peculiar velocity. The coherent velocity the redshift range [z1,z2], the resulting cross correlation component changes the galaxy number overdensity from power spectrum under the Limber’s approximation is s δg to δg ≃ δg − ∇zvz/H(z). The stochastic velocity com- χ2 −1 ponent mixes different scales and damps the power spec- Cκg(l) = 4 ng(χ)dχ (4) trum on small scales. The redshift space galaxy power Zχ1 spectrum therefore has the general form χ2 l −2 × W (χ,χs)ng(χ)P∇2 (φ−ψ)g( ,z)χ dχ 2 2 2 Zχ1 χ s 2 4 k u σv Pg (k)= Pg(k)+2u Pgθ(k)+ u Pθ(k) F 2 (1) l/χ1 H (z) W (¯χ,χs) ≃ P∇2(φ−ψ)g(k, z¯)dk 4l∆χ Zl/χ2 where u = kk/k is the cosine of the angle of the k vector (2) with respect to radial direction; Pg, Pgθ, Pθ are the real = fα(l)Pα . space galaxy power spectra of galaxies, galaxy-θ and θ, Xα respectively; σ is the 1D velocity dispersion; and F (x) v Here, χ are the comoving angular diameter distance to is a smoothing function, normalized to unity at x = 0, 1,2 redshift z1,2 andχ ¯ is the mean distance. The band power determined by the velocity probability distribution. This (2) (1) simple formula has passed tests in simulations on scales Pα of P∇2(φ−ψ)g is defined at the same k range as Pα . < In practice, we measure the band power Cκg(l, ∆l). The where δ ∼ 1 [13]. The derivation of Eq. (1) is quite general, so it should be applicable even when gravity is weighting fα(l, ∆l) is defined correspondingly. For each modified. l, only a fraction of α having fα(l, ∆l) 6= 0 contribute. s A discriminating probe of gravity.— With the The distinctive dependence of Pg on u allows for si- above measurements, one can construct an estimator multaneous determination of Pg, Pgθ and Pθ [14]. The parameters we want to determine are the band powers of C (l, ∆l) 1 ˆ κg Pgθ(k) defined such that P (k)= Pα if kα ≤ k < kα+1, EG = , (5) 2 −1 (1) (1) 3H0 a α fα(l, ∆l)Pα where k1 < k2 < · · · < kα < · · · . We denote Pα P as the band power of Pgθ. For a ki in each k bin, whose expectation value is s we have a measurement of Pg , which we denote as Pi. 2 2 The unbiased minimum variance estimator of P (1) is ∇ (φ − ψ) ∇ (φ − ψ) α hEˆGi = = . (6) Fi 2 4 2 −1 2 −1 ˆ 2 −3H a θ l 3H a βδ l P = WiPi, where Wi = (λ1 + λ2ui + λ3ui ). Here, 0 k= χ¯ ,z¯ 0 k= χ¯ ,z¯ 2σi F ≡PF (ku σ /H), σ2 is the variance of P and the three i i v i i ˆ Lagrange multipliers λα (α =1, 2, 3) is determined by The fractional error on EG is 2 2 2 2 (1),2 1 −1 2(m+n−2) Fi h∆EGi ∆C α fα∆Pα λ = (0, , 0) · A ; Amn = u . (2) ≃ + , (7) 2 i 2σ2 E2 C2 P (1) 2 Xi i G κg ( α fαPα ) P 2 2 N N Galaxy-galaxy lensing.— Weak lensing is sensitive where ∆C = [Cκg + (Cκ + Cκ )(Cg + Cg )]/(2l∆lfsky). to the convergence κ, the projected gravitational poten- N N Here, Cκ, Cκ , Cg, Cg are the power spectra of weak tial along the line of sight: lensing convergence, weak lensing shot noise, galaxy and χs galaxy shot noise, respectively, and fsky is the fractional 1 2 κ = ∇ (φ − ψ)W (χ,χs)dχ . (3) sky coverage. Errors on EG at any two adjacent bins are 2 Z0 correlated, since they always share some same k modes. However, by requiring lα/χ1 = lα+1/χ2, where l1 <l2 < · · · <lα < · · · and kα = lα/χ2, EG measurement at each 1 The determination of k relies on the angular diameter distance l bin only involves two k bins and thus only errors in DA(z) and H(z).