Skip to main content
Sandbox Physics
Redshift-space distortions

How Guzzo et al. measured growth from redshift distortions in 2008

Sandbox Physics ·

Guzzo and colleagues measured the squashing of about 6,000 galaxies near redshift 0.8 and found β=0.70±0.26\beta=0.70\pm0.26. With a bias taken from outside data, that became a growth rate f=0.91±0.36f=0.91\pm0.36, consistent with a cosmological constant in general relativity. The error bar was too large to rule out modified gravity; the lasting contribution was the proposal to use growth rates at several redshifts as a test of what drives cosmic acceleration.

This guide is pinned to arXiv:0802.1944v1, the only arXiv version, which carries the journal reference Nature 451, 541 (2008) and includes the Supplementary Information. Section names below follow that file. It assumes you have read Growth or a distance error?. Guzzo et al. 2008, arXiv:0802.1944v1

The question the paper asks

The opening paragraphs set up the problem. Dark energy and modified gravity can produce nearly the same expansion history, so distances alone cannot tell them apart. The growth of structure can, because it depends on how gravity acts on that expansion. The paper writes the growth rate as f(z)≈Ωm(z)γf(z)\approx\Omega_m(z)^\gamma, with γ≈0.55\gamma\approx0.55 for a cosmological constant in general relativity and γ≈0.68\gamma\approx0.68 for the DGP braneworld, an extra-dimensional change to gravity. Models with the same expansion but different gravity then differ in f(z)f(z).

Redshift distortions measure β=f/bL\beta=f/b_L, where bLb_L is the linear bias of the galaxies. The paper quotes an earlier 2dFGRS value at z≈0.15z\approx0.15: β=0.49±0.09\beta=0.49\pm0.09 with bL=1.0±0.1b_L=1.0\pm0.1, giving f=0.49±0.14f=0.49\pm0.14.

Reading Figure 1

Panel a is the correlation function ξ(rp,π)\xi(r_p,\pi): separation across the sky rpr_p on the horizontal axis and along the line of sight π\pi on the vertical axis. The sample is the VVDS-Wide F22 field, 4 square degrees with 5,895 redshifts between z=0.6z=0.6 and 1.21.2; the caption rounds this to about 6,000. Pixels are 1 h−1Mpc1\,h^{-1}\mathrm{Mpc} on a side and smoothed with a Gaussian before plotting. The measurement fills one quadrant and is mirrored into the other three, so do not read the symmetry as four independent measurements.

The caption names both patterns. The flattening of the outer blue and green contours is the infall; the elongation of the inner white, yellow and red contours along π\pi is the Fingers of God. Without peculiar motions the contours would be circles. The solid lines are the best-fitting model.

Panel b shows confidence regions for β\beta and the pairwise dispersion σ12\sigma_{12}. They come from 100 mock measurements: a bivariate Gaussian fitted to the spread of the mocks, then centred on the data's best fit.

The model and the error bar

The fitted model is the linear spherical-harmonic model for ξ(rp,π)\xi(r_p,\pi), convolved with an exponential distribution of pairwise velocities. It has two free parameters, β\beta and σ12\sigma_{12}. The authors report that this direct maximum-likelihood fit gave the least biased and most stable results among the methods they tested on simulations. The best fit is β=0.70\beta=0.70 and σ12=412 km s−1\sigma_{12}=412\ \mathrm{km\,s^{-1}}.

The error does not come from the formal likelihood. The authors ran the same analysis on 100 mock surveys built from the Millennium simulation, with the VVDS mask, sampling and redshift errors. The mock mean is β=0.62±0.03\beta=0.62\pm0.03 and σ12=382±12 km s−1\sigma_{12}=382\pm12\ \mathrm{km\,s^{-1}}. The Supplementary Information gives the parent simulation's true value as 0.64±0.030.64\pm0.03, so the method is not visibly biased on mocks. Marginalising over σ12\sigma_{12} gives β=0.70±0.26\beta=0.70\pm0.26.

The effective redshift is z=0.77z=0.77, the mean of the squared redshift distribution N2(z)N^2(z), chosen because ξ\xi counts pairs. The mocks show that β\beta measured this way matches β(0.77)\beta(0.77) to within 3%.

From beta to growth: the bias step

Converting β\beta to ff needs the bias, and this step uses outside data. The authors measure σ8\sigma_8 of the galaxies by counting them in random spheres, and scale the WMAP value σ8mass(0)=0.78±0.03\sigma_8^{\mathrm{mass}}(0)=0.78\pm0.03 to z=0.77z=0.77 with linear theory. The ratio gives bL=1.3±0.1b_L=1.3\pm0.1, so f=βbL=0.91±0.36f=\beta b_L=0.91\pm0.36.

You can check the arithmetic. 0.70×1.3=0.910.70\times1.3=0.91. Adding the fractional errors of β\beta and bLb_L in quadrature gives ±0.35\pm0.35, close to the quoted 0.360.36; the paper does not show its propagation, so the small difference is not explained in the text.

For comparison, a flat cosmological-constant model with Ωm0=0.25\Omega_{m0}=0.25, the cosmology of the paper's figure, has Ωm(0.77)=0.649\Omega_m(0.77)=0.649 and f≈Ωm0.55=0.79f\approx\Omega_m^{0.55}=0.79. The measurement sits about one-third of its error bar above that.

Reading Figure 2

Figure 2 plots ff against 1+z1+z. The red circle is the new point at z=0.77z=0.77. The point at z≈0.15z\approx0.15 is the 2dFGRS value. The open square at z=0.55z=0.55 comes from 2SLAQ luminous red galaxies, but the caption warns that its β\beta and bLb_L are not independent: they were fitted with consistency to the clustering at z=0z=0 imposed, which pushes the point toward the flat model.

The curves are the flat cosmological-constant model, an open model with the same matter density and no cosmological constant, two models in which dark energy couples to dark matter, and the DGP braneworld. The dotted line at f=1f=1 is a matter-only flat universe. The authors state plainly that the new point alone cannot discriminate among these models. Their stronger claim is that the three points together disfavour an open matter-only universe. A purely illustrative fit favours γ≈0.55\gamma\approx0.55 to 0.60.6, which they describe as low confidence.

The distance check in the supplement

A wrong cosmology for converting redshift to distance adds the Alcock-Paczynski distortion. The Supplementary Information tests this on the mocks: distances were recomputed for Ωm(0)\Omega_m(0) from 0.1 to 1, β\beta was remeasured, and the implied Ωm\Omega_m was fed back. The iteration converged to the true value from either side. The authors conclude that their β\beta is robust to the distance error, and add that the same result makes the Alcock-Paczynski effect hard to use at these redshifts, because the velocity distortion dominates. Compare this with the Lab's distance stretch, where the fit must carry both effects.

What later work changed

Two issues in this paper became central. First, the dispersion model. Scoccimarro showed in 2004 that the Kaiser formula holds only for kμ≪0.2 h Mpc−1k\mu\ll0.2\,h\,\mathrm{Mpc^{-1}} at redshift zero, that the density and velocity power spectra must be kept separate beyond linear order,

Ps(k,μ)=Pδδ(k)+2fμ2Pδθ(k)+f2μ4Pθθ(k),P_s(k,\mu)=P_{\delta\delta}(k)+2f\mu^2P_{\delta\theta}(k)+f^2\mu^4P_{\theta\theta}(k),

and that a constant Gaussian damping factor is "clearly an oversimplification". Scoccimarro 2004, eqs. 57, 58 and 71

Second, the bias step. Percival and White showed that the anisotropy constrains fσ8f\sigma_8 without needing the galaxy bias, which removes the outside σ8\sigma_8 and bias inputs used here. Percival and White 2009, abstract That is why later surveys report fσ8f\sigma_8.

Printed versus used

Some text extractions of the abstract show "b = 0.70 ± 0.26". The main text and Figure 1 caption make clear that the symbol is β\beta. Nothing in this guide corrects a printed equation. The quadrature check above is our own arithmetic, not the paper's.

Try it in the Lab

The Lab cannot reproduce this measurement: its synthetic galaxies sit at z=0.5z=0.5 with bias 2, and its dispersion is Gaussian per galaxy rather than exponential per pair. You can still compare scales. Set the growth to 0.910.91 and the random motion to 290 km s−1290\ \mathrm{km\,s^{-1}}, which is 412/2412/\sqrt{2}. The Lab's box holds 4.1×106 h−3Mpc34.1\times10^6\,h^{-3}\mathrm{Mpc^3}, close to the 6.35×1066.35\times10^6 of the F22 field. Across 12 such boxes the Lab's β\beta scatters by about 0.230.23, the same size as the paper's ±0.26\pm0.26, although the methods differ.

Sources

  1. Guzzo et al. 2008, arXiv:0802.1944v1
  2. Scoccimarro 2004, eqs. 57, 58 and 71
  3. Percival and White 2009, abstract

Keep reading

What this page links to directly

You are hereHow Guzzo et al. measured growth from redshift distortions in 2008

Articles

See this series on the content map →

Redshift-space distortions · Reading path

From a moving galaxy to a growth measurement

  1. ArticleWhy do galaxy maps point at us?
  2. ArticleHow a redshift moves a galaxy
  3. ArticleMultipoles of a squashed map
  4. ArticleGrowth or a distance error?
  5. ArticleMeasure multipoles from a frozen catalogue
  6. You are hereHow Guzzo et al. measured growth from redshift distortions in 2008
Something broken, confusing, or missing on this page?Tell us
What kind of feedback?

Sent with this page’s address, its Lab ID, the site language and your device type. Please leave out personal details. Feedback is kept for up to 12 months.Privacy policy