Skip to main content
Sandbox Physics
Redshift-space distortions

Multipoles of a squashed map

Sandbox Physics ·

Two numbers per scale capture most of the distortion: the monopole, the power averaged over all directions, and the quadrupole, which measures how much more power lies along the line of sight than across it. Linear infall makes both larger by fixed factors that depend only on β=f/b\beta=f/b. Random motion inside clusters lowers both and, on small scales, drives the quadrupole negative.

This chapter follows How a redshift moves a galaxy. You need polynomial integrals and the idea of an average over directions. By the end you should be able to derive the Kaiser multipoles, add a Gaussian velocity spread, and check the Lab's chart by hand.

Power as a function of angle

The previous chapter showed that each density wave in redshift space is multiplied by b+fμ2b+f\mu^2. The power spectrum is the mean squared amplitude, so

Ps(k,μ)=(b+fμ2)2P(k)=b2(1+βμ2)2P(k),β≡fb.P_s(k,\mu)=\bigl(b+f\mu^2\bigr)^2P(k)=b^2\bigl(1+\beta\mu^2\bigr)^2P(k),\qquad\beta\equiv\frac{f}{b}.

Here P(k)P(k) is the matter power spectrum and b2P(k)b^2P(k) is the galaxy power spectrum in real space. The redshift-space power depends on direction only through μ2\mu^2, the squared cosine of the angle between the wave and the line of sight. A function of μ2\mu^2 alone can be written exactly as a short sum of even Legendre polynomials.

Legendre multipoles

The Legendre polynomials are L0=1\mathcal{L}_0=1, L2=(3μ2−1)/2\mathcal{L}_2=(3\mu^2-1)/2 and L4=(35μ4−30μ2+3)/8\mathcal{L}_4=(35\mu^4-30\mu^2+3)/8. They are orthogonal on −1≤μ≤1-1\le\mu\le1, so each coefficient is found by one integral:

Pℓ(k)=2ℓ+12∫−11Ps(k,μ) Lℓ(μ) dμ.P_\ell(k)=\frac{2\ell+1}{2}\int_{-1}^{1}P_s(k,\mu)\,\mathcal{L}_\ell(\mu)\,d\mu.

The integrand is even in μ\mu, so the integral over [−1,1][-1,1] is twice the integral over [0,1][0,1]. For the monopole,

P0b2P=∫01(1+2βμ2+β2μ4) dμ=1+23β+15β2.\frac{P_0}{b^2P}=\int_0^1\bigl(1+2\beta\mu^2+\beta^2\mu^4\bigr)\,d\mu=1+\frac{2}{3}\beta+\frac{1}{5}\beta^2.

For the quadrupole you need three small integrals: ∫01L2 dμ=0\int_0^1\mathcal{L}_2\,d\mu=0, ∫01μ2L2 dμ=2/15\int_0^1\mu^2\mathcal{L}_2\,d\mu=2/15 and ∫01μ4L2 dμ=4/35\int_0^1\mu^4\mathcal{L}_2\,d\mu=4/35. Then

P2b2P=5(2β⋅215+β2⋅435)=43β+47β2.\frac{P_2}{b^2P}=5\Bigl(2\beta\cdot\frac{2}{15}+\beta^2\cdot\frac{4}{35}\Bigr)=\frac{4}{3}\beta+\frac{4}{7}\beta^2.

The same steps give P4/(b2P)=8β2/35P_4/(b^2P)=8\beta^2/35, and every higher multipole vanishes. These are Hamilton's equations 5.7. The ratio of quadrupole to monopole does not depend on the shape of P(k)P(k), which is why it became the standard way to measure β\beta. Hamilton 1998, §5.2, eqs. 5.7 and 5.8

Worked example: the Lab's default galaxies

The Lab's synthetic galaxies have bias b≈2b\approx2 on large scales and growth rate f=0.76f=0.76, so β=0.38\beta=0.38:

P0b2P=1+0.2533+0.0289=1.282,P2b2P=0.5067+0.0825=0.589.\frac{P_0}{b^2P}=1+0.2533+0.0289=1.282,\qquad\frac{P_2}{b^2P}=0.5067+0.0825=0.589.

The hexadecapole is 8×0.382/35=0.0338\times0.38^2/35=0.033, small enough that the Lab does not measure it. The Lab divides the redshift-space power by the real-space power of the same galaxies, cell by cell in wavenumber and angle, so b2Pb^2P cancels and the chart shows these ratios directly. In the default preset, the lowest bin at k=0.056 h Mpc−1k=0.056\,h\,\mathrm{Mpc^{-1}} gives a monopole of 1.2741.274 and a quadrupole of 0.5890.589. With the random motion switched off, the same bin gives 1.2921.292 and 0.5920.592. Both agree with linear theory to within the scatter of a 160 h−1Mpc160\,h^{-1}\mathrm{Mpc} box.

Adding random motion

Inside groups and clusters galaxies move fast in random directions. Give each galaxy an extra line-of-sight shift drawn from a Gaussian with standard deviation σ=(1+z)σv/H\sigma=(1+z)\sigma_v/H, where σv\sigma_v is the one-dimensional velocity dispersion. Averaging e−ikμ Δe^{-ik\mu\,\Delta} over a Gaussian shift Δ\Delta multiplies each wave amplitude by e−k2μ2σ2/2e^{-k^2\mu^2\sigma^2/2}, so the power is multiplied by its square:

Ps(k,μ)=b2(1+βμ2)2e−k2μ2σ2P(k).P_s(k,\mu)=b^2\bigl(1+\beta\mu^2\bigr)^2e^{-k^2\mu^2\sigma^2}P(k).

The course reference checks the amplitude factor with no formula from the Lab. It moves a lattice of particles by one plane wave, adds Gaussian shifts with σ=3 h−1Mpc\sigma=3\,h^{-1}\mathrm{Mpc}, and measures the wave directly. At kμ=0.157 h Mpc−1k\mu=0.157\,h\,\mathrm{Mpc^{-1}} it finds 0.89560.8956 against the predicted 0.89490.8949.

Papers usually quote the pairwise dispersion σ12\sigma_{12}, the spread in velocity difference between two galaxies. Hamilton writes the Gaussian damping as exp⁡[−(σ12kμ)2/2]\exp[-(\sigma_{12}k\mu)^2/2] and notes that a pairwise dispersion is 2\sqrt{2} larger than the single-galaxy value. So the Lab's σv\sigma_v equals σ12/2\sigma_{12}/\sqrt{2}. Many analyses use an exponential velocity distribution instead, whose damping is 1/[1+(σ12kμ)2/2]1/[1+(\sigma_{12}k\mu)^2/2]. Neither form follows from a theory of cluster dynamics; both are empirical. Hamilton 1998, §7.2, eqs. 7.2 to 7.4

How fingers change the multipoles

Without infall, β=0\beta=0, the monopole damping has a closed form. With a=k2σ2a=k^2\sigma^2,

P0b2P=∫01e−aμ2 dμ=π erf⁡(a)2a.\frac{P_0}{b^2P}=\int_0^1e^{-a\mu^2}\,d\mu=\frac{\sqrt{\pi}\,\operatorname{erf}(\sqrt{a})}{2\sqrt{a}}.

The Fingers of God only preset sets f=0f=0 and σv=450 km s−1\sigma_v=450\ \mathrm{km\,s^{-1}}, so σ=1.5×450/131.8=5.12 h−1Mpc\sigma=1.5\times450/131.8=5.12\,h^{-1}\mathrm{Mpc}. At k=0.132 h Mpc−1k=0.132\,h\,\mathrm{Mpc^{-1}}, a=0.457a=0.457 and the formula gives 0.8660.866; the Lab measures 0.8640.864. The quadrupole integral 5∫01e−aμ2L2 dμ5\int_0^1e^{-a\mu^2}\mathcal{L}_2\,d\mu evaluated numerically gives −0.252-0.252; the Lab measures −0.260-0.260. The damping removes power along the line of sight, so the quadrupole goes negative and grows in size with kk.

When infall and fingers act together, infall wins on large scales and fingers win on small ones. In the default preset the quadrupole falls from 0.5890.589 at k=0.056k=0.056 to 0.1810.181 at k=0.132 h Mpc−1k=0.132\,h\,\mathrm{Mpc^{-1}}.

Squashed in one picture, stretched in the other

The correlation function ξ\xi, the excess probability of finding a pair of galaxies at a given separation, is the Fourier transform of the power spectrum. Its multipoles carry an extra factor iℓi^\ell, so the quadrupole changes sign. A positive power-spectrum quadrupole, extra power along the line of sight, becomes a negative correlation quadrupole: contours of ξ\xi squashed along the line of sight. Hamilton 1998, §5.2, eq. 5.10

The Lab's second plot shows ξ\xi as a map of separation across the sky against separation along the line of sight, with real-space contours dashed on top. On large separations the redshift-space contours are flatter than the dashed ones. Near the centre they are pulled out vertically by the fingers.

Where the model stops working

The Gaussian factor is a phenomenological patch, and so is Kaiser's formula once structures become nonlinear. Hamilton notes that nonlinear effects reach the redshift power spectrum even on half-wavelengths as large as about 40 h−1Mpc40\,h^{-1}\mathrm{Mpc}. Scoccimarro shows that the Kaiser formula needs kμ≪0.2 h Mpc−1k\mu\ll0.2\,h\,\mathrm{Mpc^{-1}} at redshift zero, and that a constant Gaussian factor is "clearly an oversimplification" that does not suppress enough power at intermediate angles. Hamilton 1998, §7 Scoccimarro 2004, eqs. 57 and 71

The Lab therefore fits only bins with k<0.23 h Mpc−1k<0.23\,h\,\mathrm{Mpc^{-1}}, a limit set once from the 12-box test ensemble and used unchanged for every preset. Points beyond it are drawn on a shaded band for reference only.

Lab tasks

  1. Choose Coherent infall only. Read the monopole and quadrupole in the first bin and compare them with 1.2821.282 and 0.5890.589.
  2. Choose Fingers of God only. Watch the quadrupole cross zero and become more negative toward high kk.
  3. Start from the default and raise the random motion step by step. Note at which velocity the quadrupole in the third bin first drops below zero.
  4. In the correlation map, compare the outer contours with the dashed real-space ones for Coherent infall only and for Fingers of God only.

Exercises

  1. Show that the Kaiser quadrupole-to-monopole ratio is 0.45950.4595 for β=0.38\beta=0.38.
  2. What fraction of the monopole is the hexadecapole for β=1\beta=1?
  3. For σv=300 km s−1\sigma_v=300\ \mathrm{km\,s^{-1}} at z=0.5z=0.5, at what wavenumber has the line-of-sight power fallen by a factor ee?
  4. A paper quotes a pairwise dispersion of 412 km s−1412\ \mathrm{km\,s^{-1}}. What single-galaxy dispersion should you set in the Lab to compare?
Answers

For the first exercise, 0.589/1.282=0.45950.589/1.282=0.4595. Because b2Pb^2P cancels in the ratio, the result is the same at every wavenumber in the linear regime.

For the second, the monopole is 1+2/3+1/5=28/151+2/3+1/5=28/15 and the hexadecapole is 8/358/35, so the fraction is (8/35)/(28/15)=0.122(8/35)/(28/15)=0.122.

For the third, the power along the line of sight carries e−k2σ2e^{-k^2\sigma^2}, which equals e−1e^{-1} at k=1/σk=1/\sigma. With σ=1.5×300/131.8=3.41 h−1Mpc\sigma=1.5\times300/131.8=3.41\,h^{-1}\mathrm{Mpc}, that is k=0.29 h Mpc−1k=0.29\,h\,\mathrm{Mpc^{-1}}, just beyond the Lab's fit limit. The fingers still matter inside the fit range because the damping starts quadratically.

For the fourth, divide by 2\sqrt{2}: 412/1.414=291 km s−1412/1.414=291\ \mathrm{km\,s^{-1}}. The comparison is only rough, because that paper used an exponential distribution, not a Gaussian.

Sources

  1. Hamilton 1998, §5.2, eqs. 5.7 and 5.8
  2. Scoccimarro 2004, eqs. 57 and 71

Keep reading

What this page links to directly

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. You are hereMultipoles of a squashed map
  4. ArticleGrowth or a distance error?
  5. ArticleMeasure multipoles from a frozen catalogue
  6. ArticleHow 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