Redshift‑Space Distortions and Growth Rate

Introduction

The Universe is a tapestry of structures—galaxies, clusters, and filaments—woven together by gravity over billions of years. While the positions of these objects on the sky give us a two‑dimensional map, the third dimension comes from their redshifts, which encode both the smooth expansion of space and the galaxies’ own motions, known as peculiar velocities. These velocities subtly stretch and compress the observed clustering pattern when we view the cosmos in redshift space, an effect called redshift‑space distortion (RSD).

Why does this matter? Because the amplitude of peculiar velocities is directly tied to how fast matter is falling into overdensities, i.e., the logarithmic growth rate \(f \equiv d\ln D / d\ln a\) (with \(D\) the linear growth factor and \(a\) the scale factor). Measuring \(f\) across cosmic time lets us test whether gravity behaves as Einstein’s General Relativity (GR) predicts on the largest scales, or whether new physics—perhaps a modification of gravity or an exotic dark‑energy component—takes over. In the language of cosmology, RSD are the cleanest anisotropic signal that translates the invisible flow of matter into a directly observable statistic.

For a platform like Apiary, which champions bee conservation and the responsible development of self‑governing AI agents, the story of RSD is a reminder of how collective behavior—whether of galaxies or buzzing colonies—can be decoded through careful measurement and sophisticated modeling. The same statistical tools that let us read the Universe’s growth history also inspire AI systems that monitor bee populations, predict disease spread, and allocate conservation resources efficiently.

In the sections that follow we will trace the physics behind RSD, describe how anisotropic clustering yields the growth rate, explore the surveys that have turned theory into data, discuss the stringent tests of GR, and highlight the emerging role of AI in this enterprise. Along the way we’ll sprinkle concrete numbers, equations, and real‑world analogies, and we’ll link to related concepts with the slug format used throughout Apiary’s knowledge base.

The Physics of Redshift‑Space Distortions

Peculiar velocities and the mapping from real to redshift space

When a galaxy sits at comoving coordinate r, its observed redshift \(z_{\rm obs}\) receives two contributions: the cosmological redshift \(z_{\rm cos}\) from the Hubble flow and a Doppler shift from the line‑of‑sight component of its peculiar velocity v. In the distant‑observer approximation (valid for surveys covering a small solid angle), the mapping is

\[ s = r + \frac{v_{\parallel}}{aH}\,\hat{\mathbf{n}}, \]

where \(s\) is the redshift‑space position, \(v_{\parallel}\) the velocity component along the line of sight \(\hat{\mathbf{n}}\), \(a\) the scale factor, and \(H\) the Hubble parameter at the galaxy’s redshift. This mapping compresses overdense regions (where matter streams inward) and stretches underdense regions (where matter streams outward), imprinting a characteristic anisotropy on the observed clustering pattern.

The Kaiser effect

On large, linear scales (\(k \lesssim 0.1\,h\,{\rm Mpc}^{-1}\)), the coherent infall of galaxies into overdensities produces a squashing of the correlation function along the line of sight. Kaiser (1987) showed that, in Fourier space, the redshift‑space power spectrum \(P^{s}(k,\mu)\) can be written

\[ P^{s}(k,\mu) = \bigl( b + f\mu^{2} \bigr)^{2} P_{\rm m}(k), \]

where

• \(b\) is the linear bias (the ratio of galaxy to matter overdensity),
• \(f\) is the growth rate,
• \(\mu \equiv \cos\theta\) is the cosine of the angle between k and the line of sight, and
• \(P_{\rm m}(k)\) is the real‑space matter power spectrum.

The term \(f\mu^{2}\) amplifies modes that point along the line of sight, creating the characteristic anisotropic pattern that can be decomposed into Legendre multipoles (monopole, quadrupole, hexadecapole).

Finger‑of‑God (FoG) damping

On smaller, nonlinear scales (\(k \gtrsim 0.2\,h\,{\rm Mpc}^{-1}\)), random motions inside virialised halos—think galaxies orbiting within a cluster—produce an elongation of structures along the line of sight, the so‑called Finger‑of‑God effect. A common phenomenological model multiplies the Kaiser formula by a damping factor, often taken to be Lorentzian or Gaussian:

\[ D_{\rm FoG}(k,\mu,\sigma_{p}) = \frac{1}{1 + (k\mu\sigma_{p})^{2}}, \]

where \(\sigma_{p}\) is the one‑dimensional pairwise velocity dispersion (typically \(300\!-\!500\ {\rm km\,s^{-1}}\) for luminous red galaxies). Accurately separating Kaiser squashing from FoG damping is a central challenge in extracting \(f\) from data.

From Anisotropic Clustering to the Growth Rate

Multipole expansion of the correlation function

The observable two‑point statistics—either the correlation function \(\xi(s,\mu)\) or the power spectrum \(P^{s}(k,\mu)\)—are anisotropic. By expanding in Legendre polynomials \(L_{\ell}(\mu)\),

\[ \xi_{\ell}(s) = \frac{2\ell+1}{2}\int_{-1}^{1}\!\! d\mu\; \xi(s,\mu)\,L_{\ell}(\mu), \]

we isolate the monopole (\(\ell=0\)), quadrupole (\(\ell=2\)), and hexadecapole (\(\ell=4\)). The ratio of quadrupole to monopole is especially sensitive to \(f/b\), because the monopole contains a mixture of density and velocity contributions while the quadrupole isolates the \(\mu^{2}\) term.

The observable \(f\sigma_{8}\)

Cosmologists often quote the combined parameter

\[ f\sigma_{8}(z) \equiv f(z)\,\sigma_{8}(z), \]

where \(\sigma_{8}(z)=\sigma_{8,0}\,D(z)\) is the rms matter fluctuation in spheres of radius \(8\,h^{-1}{\rm Mpc}\) at redshift \(z\). This combination is directly measurable from the amplitude of the quadrupole without needing an external prior on the bias \(b\). For example, the BOSS CMASS sample (median \(z\approx0.57\)) yielded \(f\sigma_{8}=0.426\pm0.029\) (Alam et al. 2017), a 7% precision that already constrains deviations from GR at the 10% level.

Modeling pipelines

Modern analyses typically follow a pipeline:

1. Measure the multipoles of \(\xi(s)\) or \(P(k)\) from a galaxy catalogue, applying weights to correct for fiber collisions, redshift failures, and survey geometry.
2. Model the anisotropic signal using perturbation theory (e.g., the TNS model, Taruya et al. 2010) or effective field theory (EFT) approaches, incorporating bias terms up to second order and a FoG damping term.
3. Fit the model to data with a likelihood that accounts for the full covariance matrix (often estimated from thousands of mock catalogues).
4. Extract posterior distributions for \(f\sigma_{8}\), \(b\sigma_{8}\), and nuisance parameters such as \(\sigma_{p}\).

The robustness of the result hinges on the accuracy of the theoretical template; recent work shows that EFT can reach sub‑percent bias up to \(k\approx0.3\,h\,{\rm Mpc}^{-1}\) for upcoming surveys like DESI.

Survey Techniques and Data: Past, Present, Future

BOSS and eBOSS: the legacy of SDSS‑III/IV

The Baryon Oscillation Spectroscopic Survey (BOSS) measured redshifts for 1.5 million galaxies over \(10,000\ {\rm deg}^{2}\) (DR12), reaching a median redshift \(z\simeq0.57\). Its RSD analysis produced the first sub‑10% measurements of \(f\sigma_{8}\) across three redshift bins (0.38, 0.51, 0.61).

The extended BOSS (eBOSS) added quasars (up to \(z\approx2.2\)) and emission‑line galaxies (ELGs), expanding the redshift lever arm. eBOSS’s quasar sample delivered \(f\sigma_{8}=0.365\pm0.048\) at \(z=1.52\), the highest‑redshift RSD measurement to date (Zarrouk et al. 2020).

DESI: a ten‑fold increase in statistical power

The Dark Energy Spectroscopic Instrument (DESI) began operations in 2024 and aims to obtain spectra for 35 million galaxies and quasars over 14,000 deg². Its target classes include:

DESI’s forecasted precision on \(f\sigma_{8}\) is ≈1% per redshift bin, enough to discriminate between GR (\(\gamma\approx0.55\)) and many modified‑gravity models (\(\gamma\approx0.68\)) at > 5σ.

Euclid and the Roman Space Telescope: space‑based RSD

The European Space Agency’s Euclid mission (launch 2027) will perform a spectroscopic survey of 15 million Hα‑emitting galaxies over \(0.9<z<1.8\). Its slitless spectroscopy yields precise redshifts (σ_z/(1+z)≈0.001) and, combined with weak‑lensing shear, will tighten the growth‑rate constraints to ≈0.5% on scales up to \(k=0.2\,h\,{\rm Mpc}^{-1}\).

NASA’s Roman Space Telescope will complement Euclid with a high‑density Hα sample over a smaller area, enabling cross‑checks of systematic effects such as fiber‑assignment bias (which has no analogue in slitless spectroscopy).

Ground‑based synergy: LSST and CMB‑S4

The Vera C. Rubin Observatory’s Legacy Survey of Space and Time (LSST) will deliver deep photometric redshifts for billions of galaxies, providing an independent probe of photometric RSD via clustering‑redshift techniques. Meanwhile, the upcoming CMB‑S4 experiment will map the kinetic Sunyaev–Zel’dovich (kSZ) effect, offering a direct measurement of peculiar velocities that can be cross‑correlated with galaxy RSD to break degeneracies between bias and growth.

Testing General Relativity on Cosmic Scales

The growth index \(\gamma\)

A convenient phenomenological parameterization of the growth rate is

\[ f(z) = \Omega_{m}(z)^{\gamma}, \]

where \(\Omega_{m}(z) = \frac{\Omega_{m,0}(1+z)^{3}}{E^{2}(z)}\) and \(E(z)=H(z)/H_{0}\). In GR with a cosmological constant, \(\gamma\simeq0.55\) (to within 1% for \(\Lambda\)CDM). Modified‑gravity theories (e.g., \(f(R)\), DGP) predict larger values, typically \(\gamma\approx0.68\).

By fitting the measured \(f\sigma_{8}(z)\) from multiple redshift bins, one can infer \(\gamma\). Current combined analyses (BOSS + eBOSS + VIPERS) give \(\gamma = 0.558 \pm 0.045\) (Alam et al. 2022), fully consistent with GR. The upcoming DESI and Euclid data are projected to shrink the uncertainty to ≈0.015, a regime where subtle deviations—perhaps from a time‑varying dark‑energy equation of state—could become detectable.

Model‑independent consistency tests

Beyond the \(\gamma\) parametrization, the growth–geometry consistency test compares the expansion history inferred from distance indicators (e.g., supernovae, BAO) with the growth history from RSD. In GR, the two are linked through the Einstein equations; a mismatch signals new physics. Recent work using the \(E_{G}\) statistic (combining RSD, weak lensing, and galaxy clustering) finds \(E_{G}=0.43\pm0.07\) at \(z=0.6\), again consistent with GR’s prediction of 0.40.

Constraints on specific modified‑gravity models

• \(f(R)\) gravity: The scalaron field introduces a scale‑dependent enhancement of the growth rate. RSD analyses from BOSS place an upper bound \(|f_{R0}|<10^{-5}\) (95% CL).
• Dvali‑Gabadadze‑Porrati (DGP) braneworld: Predicts \(\gamma\approx0.68\). Combined RSD and supernova data rule out the self‑accelerating branch at > 4σ.
• Early‑dark‑energy (EDE) models: Slightly increase \(\Omega_{m}(z)\) at high redshift, leading to a modest boost in \(f\). Current RSD constraints limit the EDE fraction to < 2% of the total energy density at recombination.

Systematics and Modeling Challenges

Nonlinear bias and higher‑order terms

Linear bias (\(b\)) suffices on large scales, but on \(k\gtrsim0.1\,h\,{\rm Mpc}^{-1}\) the relationship between galaxy and matter overdensities becomes nonlinear. The bias expansion includes terms such as \(b_{2}\delta^{2}\) and tidal‑bias \(b_{s^{2}}s^{2}\). Ignoring them can shift the inferred \(f\sigma_{8}\) by 5–10%. Modern EFT‑based analyses marginalize over these coefficients with priors from N‑body simulations.

Alcock‑Paczynski (AP) effect

If the assumed cosmology used to convert redshifts to distances differs from the true one, the observed clustering appears distorted even without peculiar velocities. The AP parameters

\[ \alpha_{\parallel} = \frac{H_{\rm fid}(z)}{H(z)},\qquad \alpha_{\perp} = \frac{D_{A}(z)}{D_{A,{\rm fid}}(z)}, \]

scale the line‑of‑sight and transverse separations, respectively. Simultaneously fitting \((\alpha_{\parallel},\alpha_{\perp})\) together with \(f\sigma_{8}\) mitigates bias, but degeneracies remain; a 1% error in \(\alpha_{\parallel}\) translates to a comparable error in \(f\).

Survey selection and fiber collisions

Multi‑object spectrographs cannot place fibers arbitrarily close; typical minimum separations are 62 arcsec for BOSS. This “fiber‑collision” effect preferentially removes close pairs, suppressing the small‑scale quadrupole. The standard correction—upweighting the nearest neighbor—works to ≈2% accuracy for BOSS, but DESI’s higher density demands more sophisticated pairwise‑inverse‑probability weighting.

Redshift errors

Spectroscopic redshift uncertainties (σ_z ≈ 30 km s⁻¹ for LRGs, ≈ 70 km s⁻¹ for ELGs) add an extra Gaussian damping term similar to FoG. For ELGs, the damping can reduce the quadrupole amplitude by ≈5% at \(k=0.2\,h\,{\rm Mpc}^{-1}\) if unaccounted for.

The Role of AI and Machine Learning in RSD Analysis

Emulators for rapid theory predictions

Computing the full EFT‑RSD model for each point in a high‑dimensional parameter space is computationally expensive. Neural‑network emulators trained on a sparse grid of N‑body simulations can predict the multipole spectra to sub‑percent accuracy in milliseconds. The CosmoPower framework (Moser et al. 2022) has already been integrated into the DESI likelihood pipeline, reducing wall‑clock time from hours to seconds.

Self‑governing agents for survey optimization

Apiary’s research into self‑governing AI agents—software entities that negotiate resource allocation without central oversight—finds a natural application in adaptive survey strategies. An agent could monitor the real‑time RSD signal-to-noise across the sky and re‑prioritize fiber assignments to maximize the Fisher information on \(f\sigma_{8}\). Early simulations suggest a 10% improvement in overall growth‑rate precision compared to static tiling.

Anomaly detection in massive catalogues

With billions of spectra, human vetting of outliers is impossible. Convolutional autoencoders trained on the spectral shape can flag galaxies with abnormal emission‑line ratios or incorrect redshift assignments, both of which could bias RSD if left unchecked. A recent pilot on the eBOSS ELG sample reduced the systematic shift in \(f\sigma_{8}\) from 0.03 to < 0.01 after cleaning flagged objects.

Cross‑disciplinary lessons from bee‑colony modeling

Ecologists model bee foraging using agent‑based simulations that capture collective decision‑making. The same stochastic, locally interacting agents underpin halo‑occupation distribution (HOD) models for galaxies. Techniques such as reinforcement learning—used to evolve optimal foraging strategies—are now being explored to train HOD parameters that best reproduce observed RSD multipoles, offering a fresh perspective on bias modeling.

Analogies from Nature: Collective Behavior in Bee Colonies

A bee colony is a self‑organized system where thousands of individuals follow simple interaction rules—pheromone trails, waggle dances—to achieve efficient foraging and thermoregulation. The emergent pattern—clusters of bees around the queen, radial gradients of temperature—mirrors the cosmic web’s filamentary structure.

• Clustering strength: In cosmology, the bias parameter quantifies how strongly galaxies trace the underlying matter field. In a hive, the density of foragers around a rich flower patch is analogous to a high‑bias tracer. Both systems can be described by a two‑point correlation function, albeit with different physical units.

• Information flow: Peculiar velocities convey the gravitational “communication” between overdensities. Bees transmit information via the waggle dance; the speed and direction of the dance encode the location of resources, much like velocities encode the growth of structure.

• Growth measurement: The population growth rate of a hive (births minus deaths) can be measured by counting brood cells over time. Similarly, the cosmological logarithmic growth factor \(f\) quantifies how quickly matter perturbations amplify. Both are dimensionless