Download Linear Perturbation Theory - Lecture Notes | ASTR 596 and more Study notes Astronomy in PDF only on Docsity! Linear perturbation theory – 4 Finally, combine the continuity and Euler equations and use the Poisson equation to eliminate , yielding This is the cosmological (matterdominated) version of the Jeans perturbation equation. As before, we can take a spatial Fourier transform (comoving) to obtain Again we have instability growth if the RHS is > 0; but the critical wavenumber k J now changes with time: Modes with k > k J oscillate (sound waves); modes with k < k J are unstable. We are interested in unstable modes... so drop the pressure term... ̈2 ȧ a ̇− cs 2 a2 ∇ 2−4G = 0 ̈k2 ȧ a ̇k = 4G − cs2 k2a2 k (matter domination, subhorizonscale) k J = 2 cs G a2 = 2cs G0a Linear perturbation theory – 5 If we consider only unstable modes and drop the pressure term, we can use the resulting equation for all nonrelativistic matter on subhorizon scales, even during the radiationdominated era: Consider the case 0 = 1. We have Matter domination: Radiation domination: Radiation Jeans length is comparable to the horizon size, so treat it as smooth. The Universe expands too rapidly for dark matter fluctuations to grow. For baryons, radiation pressure keeps subhorizonscale fluctuations from growing. ̈k2 ȧ a ̇k = 4G k ̈k 4 3 t ̇k = 2 3 t2 k ⇒ k ∝ t 2/3 (growing) t−1 (decaying) a ∝ t 2/3 , ∝ a−3 a ∝ t 1/2 , m ∝ a −3 , r ∝ a −4 ̈k 1 t ̇k ≈ 0 ⇒ k ∝ ln t , constants Evolution of important proper length scales ( 0 = 1) log r log t ho riz on In fl at io n Radiation dominated Matter dominated Vacuum dominated Jeans length k 2 ??? Recombination k 1 About the size of a globular cluster About the size of a cluster of galaxies About the size of a galaxy ~ 1 Horizon crossing Spherical collapse model Collapse of a uniform, spherical perturbation (“top hat”) can be solved analytically. Consider: 2 and r 2 are chosen so that average density inside r 2 is equal to . The overdense region behaves like a portion of a closed universe model (assuming 1 > crit ). Outside r 2 the universe behaves like an unperturbed model. r (r) r 1 r 2 _ 1 2 r = {1 r r 12 = −1r1/r231−r1 /r 23 r1 r r 2 _ Spherical collapse model – 2 If we take the background cosmology to have 0 = 1, then any overdensity 1 will expand to a maximum size and then recollapse. Using the Friedmann equation solution for 0 > 1, we have at the epoch of maximum expansion where initial values (H i , i ) are taken at an early epoch a i . If we take this time to be very early, then (k = curvature) and amax a i = i i−1 H i tmax = 2 i i−1 3 /2 ≃ 2 amaxai 3 /2 max = amaxai 3 /2 i = 3 32 G tmax 2 1t i ≃ ti ≫ k /ai 2 , i ≃ 1 max ≡ 1tmax tmax −1= 92 16 −1≈ 4.55 Zel'dovich pancakes – 2 The deformation (strain) tensor is The principal axes of the deformation are determined by ∂p/∂q; because the flow is irrotational, D is symmetric in this coordinate system: From conservation of mass we can obtain the density : Notice that the density becomes infinite when any one of the terms in parentheses becomes zero. For random perturbations, the chance that any two of (, , ) are the same, or that all are the same, is vanishing – hence collapse should occur preferentially along one direction (the direction depends on the local pert. field). Dij = ∂r i ∂ q j = a t ija t Dt ∂ pi ∂ q j D = [a−a D 0 00 a−a D 00 0 a−a Dt ] a31− D1− D1− D = a 3 Zel'dovich pancakes – 4 FLASH solution to the Zel'dovich pancake problem (Zel'dovich 1970; Anninos & Norman 1994) in a flat Universe ( 0 = 1) with Hubble constant H 0 = 50 km s–1 Mpc–1 and comoving wavelength = 10 Mpc at redshift z = 0. Left Dark matter density profile (top) and phase plot (bottom). Solid curves show darkmatteronly case; dashed curves show case with 10% gas fraction. Right Gas number density, temperature, and velocity profiles. Solid curves show gasonly case; dashed curves show case with 10% gas fraction. A uniform onedimensional mesh with 8,192 zones was used for the gasdynamics and potential solver, and 65,536 particles were used to represent the dark matter. Intermission Supercomoving coordinates — 3
The Poisson equation for the comoving potential is
4nG
Vb = —s (05 + Pam) — (Pg + Pam)|
The equations of motion for particles (dark matter, stars, etc.) are
Xam _~yv
dt dm
dVam 2 vam —_vV¢
And of course the Friedmann equation can be solved numerically with an ODE
integrator.
ay? Qm % OQ
w(t) = (7) = Hi UR Get)
The power spectrum The power spectrum of density fluctuations is defined as where k is the Fourier transform of the density perturbation field. Angle brackets denote an ensemble average. Notice that in the linear regime we have In general we could have P be a function of the amplitude and direction of k, but in an isotropic universe P must be a function of the amplitude k only. The dimensionless power spectrum is defined as where V is the normalization volume (arbitrary). Pk ≡ 〈∣k∣ 2 〉 2k ≡ V 23 4 k 3 P k k ∝ D z ⇒ P k ∝ D z 2 Gaussian random fields For a Gaussian random field, the power in perturbation modes with comoving wavenumber k is distributed according to the Rayleigh distribution: The complex phase of k is uniformly distributed in [0, 2). A realization of a given power spectrum in a finite volume displays fluctuations in the actual power about P(k) due to the finite number of modes with a given wavenumber k (cosmic variance). The largest variance comes at small k, where we have the smallest number of samples. Probability that ∣k∣ 2 X is exp [−X 2/P k ] M. Tegmark Initializing cosmological simulations (uniform meshes) 1. Compute the Fourier transform of the overdensity field, k =| k |exp(i k ). For each kspace zone pqr, an exponential deviate. a uniform deviate in [0,1). Note 1: must have since (x) is realvalued. Note 2: usually choose initial redshift z so that max[(x)] < 1. 2. Inverse Fourier transform to get the realspace density fluctuation ijk = (x ijk ). 3. To get the velocity field, use and the curlfree character of v: then inverse Fourier transform to get v ijk = v(x ijk ). ∣ pqr∣= D z Pk pqr , z=0 pqr = 2 N−p , N−q , N− r = pqr , N−p , N−q , N−r =−pqr ∇⋅v =−̇ v pqr = i kpqr k pqr 2 Ḋ D pqr Initializing cosmological simulations (particles) 1. Take unperturbed positions q to lie on a grid: 2. Compute the Fourier transform of the velocity potential and velocity v: The pqr are computed as for gridbased initialization. 3. Inverse Fourier transform to get the particle velocities. The displaced particle positions are then q ijk = i x , j y , kz v = ∇ ⇒ v pqr = i kpqr pqr ∇⋅v =−̇ ⇒ ∇ 2=−̇ pqr = ̇pqr k pqr 2 vpqr = i kpqr k pqr 2 ̇ pqr . x ijk = qijk D Ḋ v ijk Zel'dovich approximation example
Dark matter particle positions
Mesh gas overdensities
IDL 32
(displacements multiplied by 7)