The DESI DR1 Peculiar Velocity Survey: growth rate measurements from galaxy and momentum correlation functions ∗ E-mail: ryan.turner@anu.edu.au R. J. Turner, 1,∗ C. Blake, 1 F. Qin, 2 J. Aguilar, 3 S. Ahlen, 4 A. J. Amsellem, 5 J. Bautista, 2 S. BenZvi, 6 D. Bianchi, 7,8 D. Brooks, 9 A. Carr, 10 E. Chaussidon, 3 T. Claybaugh, 3 A. Cuceu, 3 A. de la Macorra, 11 P. Doel, 9 K. Douglass, 6 S. Ferraro, 3,12 A. Font-Ribera, 13 J. E. Forero-Romero, 14,15 E. Gaztañaga, 16,17,18 S. Gontcho A Gontcho, 3,19 G. Gutierrez, 20 J. Guy, 3 H. K. Herrera-Alcantar, 21,22 K. Honscheid, 23,24,25 C. Howlett, 26 D. Huterer, 27,28 M. Ishak, 29 R. Joyce, 30 R. Kehoe, 31 A. G. Kim, 3 D. Kirkby, 32 A. Kremin, 3 O. Lahav, 9 Y. Lai, 26,33 C. Lamman, 25 M. Landriau, 3 L. Le Guillou, 34 A. Leauthaud, 35,36 M. E. Levi, 3 M. Manera, 37,13 A. Meisner, 30 R. Miquel, 38,13 J. Moustakas, 39 A. Muñoz-Gutiérrez, 11 S. Nadathur, 17 N. Palanque-Delabrouille, 22,3 W. J. Percival, 40,41,42 C. Poppett, 3,43,12 F. Prada, 44 I. Pérez-Ràfols, 45 C. Ross, 26 G. Rossi, 46 K. Said, 26 E. Sanchez, 47 D. Schlegel, 3 M. Schubnell, 27,28 J. Silber, 3 D. Sprayberry, 30 G. Tarlé, 28 B. A. Weaver, 30 P. Zarrouk, 34 and H. Zou 48 (Affiliations can be found after the references) Abstract Joint analysis of the local peculiar velocity and galaxy density fields offers a promising route to testing cosmological models of gravity. We present a measurement of the normalised growth rate of structure, f ​ σ 8 f\sigma_{8} , from the two-point correlations of velocity and density tracers from the DESI DR1 Peculiar Velocity and Bright Galaxy Surveys, the largest catalogues of their kind assembled to date. We fit the two-point correlation measurements with non-linear correlation function models, constructed from density and momentum power spectra generated using 1-loop Eulerian perturbation theory, and validate our methodology using representative mock catalogues. We find f ​ σ 8 = 0.391 − 0.081 + 0.080 f\sigma_{8}=0.391^{+0.080}_{-0.081} , consistent to within 1 ​ σ 1\sigma with accompanying analyses of the same datasets using power spectrum and maximum-likelihood fields methods. Combining these growth rate results from different methods including appropriate correlations, we find a consensus determination f ​ σ 8 ​ ( z = 0.07 ) = 0.4497 ± 0.0548 f\sigma_{8}(z=0.07)=0.4497\pm 0.0548 , consistent with predictions from Planck + Λ +\Lambda CDM cosmology. Jointly fitting to this consensus low-redshift growth rate and the DESI DR1 full-shape clustering dataset, we measure gravitational growth index γ L = 0.580 − 0.110 + 0.110 \gamma_{\rm L}=0.580^{+0.110}_{-0.110} , consistent with the prediction of general relativity. Keywords: Cosmology, Large-Scale Structure, Peculiar Velocities 1 Introduction The field of cosmology is currently grappling with several fundamental ‘tensions’ that have the potential to change our understanding of the universe ( Riess et al. 2016 ; Leauthaud et al. 2017 ; Joudaki et al. 2018 ; Di Valentino et al. 2021 ; Abdalla et al. 2022 , e.g.,) . It is often suggested that these problems will be resolved once newer, larger datasets become available, yet these issues seem to persist. Indeed, supernova measurements from the Dark Energy Survey ( DES Collaboration et al. 2024 , DES;) and recent Baryon Acoustic Oscillations (BAO) measurements from the Dark Energy Spectroscopic Instrument ( DESI Collaboration et al. 2016a ; DESI Collaboration et al. 2022 , DESI;) survey appear to favour a time-varying model of dark energy over the commonly-accepted cosmological constant ( DESI Collaboration et al. 2025c ; DESI Collaboration et al. 2025a ; Ishak et al. 2025 ) . The local universe is a powerful laboratory for testing this dark energy component, which dominates the local cosmic dynamics. Whilst the local expansion is chiefly sensitive to the Hubble constant, and the local volume is insufficient for accurate galaxy clustering measurements, the local motions of galaxies nonetheless allow accurate tests of gravitational physics. These so-called ‘peculiar velocities’ are induced in galaxies by the gravitational influence of nearby large-scale structure, causing departures from their expected recessional velocities. These effects modify the cosmological redshift of a galaxy as: (1+z)=(1+\bar{z})\left(1+\frac{v_{p}}{c}\right)\,, (1) where z is the measured redshift, \bar{z} is the cosmological redshift, and v_{p} is the peculiar velocity ( Davis & Scrimgeour 2014 , for a discussion of fundamentals, see) . As part of its 8-year observing program, DESI is assembling the best currently-existing spectroscopic datasets of the local Universe, including a dedicated survey of peculiar velocities. This survey is expected to obtain \sim 180{,}000 peculiar velocities by the end of the observing program ( Saulder et al. 2023 ) , roughly five times larger than the previous largest such homogeneously-selected catalogue, the Sloan Digital Sky Survey Peculiar Velocity Survey ( Howlett et al. 2022 , SDSS PV;) . This paper is one of a set describing the analysis of the first year of data from the DESI Peculiar Velocity survey, accompanying Data Release 1 ( DESI Collaboration et al. 2025b , DR1,) . In this work we will describe the measurement of the growth rate of structure from galaxy and momentum (density-weighted velocity) correlation function statistics. The growth rate of structure, f , describes the logarithmic growth of density perturbations D , with respect to the cosmic scale factor a , and can also be described by an empirical fitting formula, f\equiv\frac{d\ln D(a)}{d\ln a}\simeq\Omega_{m}(a)^{\gamma}\,, (2) where \Omega_{m}(a) is the time-dependent matter density parameter, and \gamma is the growth index as defined by Linder 2005 , where in general relativity, \gamma_{GR}=0.55 . This empirical relation is precise even when extended to cosmologies with dynamical dark energy ( Dodelson & Schmidt 2020 ) . Cosmological models that share an expansion history, but have differing gravitational theories, will find different values for \gamma . Since the amplitude of peculiar velocities is also driven by the amplitude of density fluctuations in the Universe, momentum correlations probe (at leading order) the combined parameter f\sigma_{8} , the normalised growth rate of large-scale structure. The parameter \sigma_{8} describes the r.m.s. amplitude of linear mass fluctuations on scales of R_{8}=8\,h^{-1} Mpc, measured at the present time z=0 , \sigma_{8}^{2}=\frac{1}{2\pi^{2}}\int_{0}^{\infty}dk\,k^{2}\,P(k)\,|W(kR_{8})|^{2}\,, (3) where P(k) is the linear matter power spectrum measured at the present time and W(kR_{8}) is the Fourier transform of the spherical top hat window function with radius R_{8} . To directly measure peculiar velocities we require redshift-independent distance measurements. Measuring \bar{z} in this way, in conjunction with accurate spectroscopic measurements of the observed redshift z , enables us to infer v_{p} directly from Equation 1 . The DESI Peculiar Velocity survey uses the Fundamental Plane Relation ( Djorgovski & Davis 1987 ; Dressler et al. 1987 , FP;) and the Tully-Fisher Relation ( Tully & Fisher 1977 , TF;) to measure redshift-independent distances. For a review of peculiar velocity cosmology, see Turner 2024 . Peculiar velocities also distort the redshift-space positions of galaxies, and hence induce anisotropies in the galaxy clustering signal as a function of the angle to the line-of-sight ( Kaiser 1984 ; Kaiser 1987 ) , a phenomenon known as redshift-space distortions (RSDs). Growth rate information can be extracted from RSDs by measuring the clustering signal in redshift space. Directly measured peculiar velocities and statistically inferred peculiar velocities are complementary probes. The Fourier modes of the momentum and density fields are related by \tilde{v}(k)\propto\tilde{\delta}(k)/k , causing an increase in signal-to-noise for velocities at larger scales (smaller k ), such that peculiar velocities are organised in ‘bulk flows’. The momentum power spectrum hence traces the underlying matter power spectrum on large scales, where we expect to see evidence for modified gravity models before they are screened, whilst the signal-to-noise of RSD is dominated by smaller-scale modes ( Koda et al. 2014 ) . In recent years, significant effort has been invested in producing methods for extracting cosmological information from peculiar velocities. Under the assumption that the momentum and galaxy density fields are Gaussian, we can use the two-point summary statistics between velocity and density tracers to measure the growth rate; this has been done with the Fourier-space power spectra ( Park 2000 ; Howlett et al. 2017b ; Qin et al. 2019 ; Qin et al. 2025 ) and the configuration-space correlation function ( Nusser 2017 ; Wang et al. 2018 ; Dupuy et al. 2019 ; Courtois et al. 2023 ; Turner et al. 2023 ; Lyall et al. 2024 ) . Maximum likelihood methods that analytically model the covariance between the velocity and galaxy density fields, assuming both fields are correlated samples drawn from multivariate Gaussians, have also been used to constrain the growth rate by Johnson et al. 2014 ; Huterer et al. 2017 ; Adams & Blake 2020 , and Lai et al. 2023 . Reconstruction techniques, where the velocity field is inferred from the measured galaxy density field and then directly compared to discrete velocity measurements point-by-point, are another viable method of constraining the growth rate ( Davis et al. 2011 ; Turnbull et al. 2012 ; Carrick et al. 2015 ; Said et al. 2020 ; Lilow & Nusser 2021 ; Qin et al. 2023 ; Boubel et al. 2024 ) . We refer readers to Strauss & Willick 1995 for a historical review of science achievable with the local galaxy density and peculiar velocity fields. A series of companion papers present the DESI DR1 PV datasets, and complementary cosmological analyses. Measurements of f\sigma_{8} from the power spectra are described by Qin et al. in prep. 2025 , and results from the maximum-likelihood method are described by Lai et al. in prep. 2025 . The datasets themselves are described by Ross et al. in prep. 2025 who cover the DESI DR1 Fundamental Plane catalogue, and Douglass et al. in prep. 2025 who present the DESI DR1 Tully-Fisher catalogue. The zero-point calibration for both catalogues, as well as constraints on the Hubble constant H_{0} , are described by Carr et al. in prep. 2025 . The galaxy and velocity mocks produced for this analysis are described by Bautista et al. in prep. 2025 . We will refer readers to these papers as appropriate. Our paper is structured as follows: in Section 2 we describe the DESI DR1 and mock datasets we utilise in this work. In Section 3 we describe the construction of the models and estimators used in our analysis, and the methodology we use to fit our models to the data. We present our fitting results on the mocks and data, and discuss our findings, in Section 4 . Finally, we conclude and propose ideas for future work in Section 5 . 2 The Dark Energy Spectroscopic Instrument The Dark Energy Spectroscopic Instrument is a multi-fibre spectrograph installed on the 4m Mayall Telescope at the Kitt Peak National Observatory, Arizona, USA. The DESI survey is a spectroscopic survey designed to map the large-scale structure of the universe over a wide redshift range and substantial cosmological volume, with the goal of measuring the expansion history of the universe as well as the growth rate of large-scale structure ( Levi et al. 2013 ; DESI Collaboration et al. 2016a ; DESI Collaboration et al. 2016b ) . The DESI Survey is planned to span eight years, during which it will observe more than 60 million galaxies and quasars over a footprint of 17{,}000 square degrees. The survey is internally divided into four main target classes: the Bright Galaxy Survey (BGS), the Luminous Red Galaxy Survey (LRG), the Emission Line Galaxy Survey (ELG), and the Quasar Survey (QSO), where targets are all selected from the DESI Legacy optical imaging surveys ( Dey et al. 2019 ) . The DESI focal plane covers 8 square degrees and contains 5000 optical fibers ( Poppett et al. 2024 ) , which can be individually placed onto targets in each observational field using a robotic positioner ( Silber et al. 2023 ) and an optical corrector ( Miller et al. 2024 ) . In each exposure, the observed targets are measured across the wavelength range 3600-9800 Å and the spectral data are processed by Redrock 1 1 1 https://github.com/desihub/redrock , the DESI spectroscopic pipeline ( Guy et al. 2023 ) . The overall observing strategy is summarised by Schlafly et al. 2023 . DESI has issued an Early Data Release ( DESI Collaboration et al. 2024b ) , which has been used for scientific validation of the program ( DESI Collaboration et al. 2024a ) . For this work we use a galaxy sample drawn from the publicly available DESI Data Release 1 ( DESI Collaboration et al. 2025b ) , specifically the DESI DR1 large-scale structure catalogues 2 2 2 https://data.desi.lbl.gov/doc/releases/dr1/ ( DESI Collaboration et al. 2025d ) . The velocity sample is taken from the first year of observations of the DESI Peculiar Velocity (DESI PV) survey, as described by Ross et al. in prep. 2025 and Douglass et al. in prep. 2025 . We briefly summarise these datasets below. 2.1 The Bright Galaxy Survey dataset In our study, we wish to jointly analyse the clustering statistics of the momentum and galaxy density fields. The momentum field is only well measured at redshifts z\lesssim 0.1 , as this is the range over which distance errors are not yet dominant in estimates of galaxy peculiar velocities, the intrinsic scatter in the galaxy scaling relations are not too large, and the fractional contribution of peculiar velocities to the measured redshift of sources is still significant. Thus, we use redshift data from the DESI Bright Galaxy Survey (BGS), the lowest-redshift DESI sample, to measure galaxy clustering statistics. The BGS sample is a magnitude-limited sample of galaxies with r -band magnitudes 140.05 . We illustrate this effect in Figure 1 , where we compare measurements of the \psi_{1} momentum correlation function (introduced in Section 3.2 ) from the full TF sample, the TF sample cut at z<0.05 , the FP dataset, and the combination of the FP and z -cut TF datasets. The reasons for this excess correlation (which is also seen in the other analysis methods) are yet to be understood, but in order to be conservative we only considered the z<0.05 component of the TF sample in our analysis, reducing the sample size to 2{,}934 galaxies (although we note that the TF sample does not significantly contribute to the growth rate determination at this stage of the survey, so this cut has no effect on the conclusions of our study). Applying the 4\sigma log-distance ratio cut to this subsample removes an additional 5 galaxies, leaving 2{,}929 spiral galaxies in our sample. Our final PV sample hence contains 73{,}686 FP velocities and and 2{,}929 TF velocities, for a total DR1 sample size of 76{,}615 . We choose the effective redshift of the sample to be the mean redshift of all galaxies across the clustering and velocity samples, z_{\rm eff}=0.07 . Figure 1: The \psi_{1} correlation function for different velocity datasets tested in our analysis: the FP sample (black points), the TF sample (red points), the TF sample with a z<0.05 cut (green points), and the combined sample (blue points). The solid line is the fiducial model. The original TF sample shows a high amplitude of correlation, leading us to apply a redshift cut z<0.05 for the cosmological analysis. 2.3 Mocks We created simulated clustering and peculiar velocity catalogues closely matching the BGS and PV datasets used in this analysis. These mocks enable us to perform robust tests of our methodology before applying it to the DR1 data, and are fully described by Bautista et al. in prep. 2025 . The mocks are produced from 25 realisations of the AbacusSummit simulation ( Garrison et al. 2021 ; Maksimova et al. 2021 ) , where each realisation is subsampled to produce 27 independent mock datasets. Hence we use 675 clustering and velocity mocks, which we sub-sample to match the redshift distribution and angular selection of the real datasets, and apply peculiar velocity noise matching the FP and TF samples. The AbacusSummit simulations are produced with a \Lambda CDM cosmology with matter density \Omega_{m}=0.3153 , baryon density \Omega_{b}=0.0493 , Hubble parameter h=0.6736 , normalisation \sigma_{8}=0.8114 , and spectral index n_{s}=0.9649 ( Planck Collaboration et al. 2020 , the cosmology favoured by) . Our mocks are based on snapshots generated at a redshift of z=0.2 , with a fiducial value of the normalised growth rate f\sigma_{8}=0.462 . We note that this is different to the effective redshift of the DR1 dataset, but this is the lowest-redshift snapshot that was fit with a halo occupaton distribution (HOD) model. In future work we would like to re-fit the HOD parameters to the z = 0.1 AbacusSummit snapshot, which is the lowest redshift snapshot available to us. 3 Fitting procedure 3.1 Models In this section, we describe the models we adopt for fitting the galaxy and momentum auto- and cross-correlations. We utilise the 1-loop Eulerian standard perturbation theory approach of Qin et al. in prep. 2025 to generate non-linear models of the density and momentum power spectra, which we transform to non-linear correlation function models. We will not describe in detail how these power spectra models are produced, and instead direct interested readers to the accompanying paper for a thorough description of the modelling. Here, we describe the conversion of the power spectra to the correlation function models used in our analysis. We utilise five correlation functions in total: two momentum auto-correlations, the monopole and quadrupole of galaxy clustering, and the cross-correlation dipole. 3.1.1 Momentum auto-correlation model We start by describing our models for the momentum auto-correlation functions, \psi_{1} and \psi_{2} ( Gorski et al. 1989 ) . These take the form, \psi_{1}(r)=A(r)\,\Psi_{\parallel}(r)+[1-A(r)]\,\Psi_{\perp}(r)\,, (6) \psi_{2}(r)=B(r)\,\Psi_{\parallel}(r)+[1-B(r)]\,\Psi_{\perp}(r)\,, (7) where \Psi_{\perp}(r) and \Psi_{\parallel}(r) describe the transverse and radial correlations of the three-dimensional peculiar velocity field, and the terms A(r) and B(r) represent the geometry of the survey. It has been shown that the radial velocity correlation estimators \psi_{1} and \psi_{2} contain all of the information within the 3D velocity tensor ( Gorski 1988 ; Wang et al. 2018 ; Blake & Turner 2024 ) , under the assumption of an irrotational velocity field. The functions \Psi_{\perp}(r) and \Psi_{\parallel}(r) are given by, \Psi_{\perp}(r)=\frac{1}{2\pi^{2}}\int dk\,k^{2}\,P_{v}(k)\,\left[j_{0}(kr)-2\frac{j_{1}(kr)}{kr}\right]\,, (8) \Psi_{\parallel}(r)=\frac{1}{2\pi^{2}}\int dk\,k^{2}\,P_{v}(k)\,\frac{j_{1}(kr)}{kr}\,, (9) where j_{n}(x) are the spherical Bessel functions, and we set P_{v}(k)=3P^{0}_{pp}(k) , where P^{0}_{pp}(k) is the monopole of the line-of-sight momentum power spectrum produced by the code of Qin et al. in prep. 2025 . The factor of 3 can be readily understood if we consider the leading-order model for the angle-dependent line-of-sight momentum power spectrum, P_{pp}(k,\mu)=\left(\frac{aHf\mu}{k}\right)^{2}\,P_{L}(k)\,, (10) where P_{L}(k) is the linear matter power spectrum. The monopole of Equation 10 is obtained by integrating over \mu , ranging from 0 to 1, P^{0}_{pp}(k)=\frac{1}{3}\left(\frac{aHf}{k}\right)^{2}P_{L}(k)=\frac{1}{3}P_{v}(k)\,, (11) where P_{v}(k) is the velocity power spectrum that appears in Equation 8 and Equation 9 . To convert from power spectrum multipoles to correlation function models, we utilise the python implementation of the FFTLog algorithm within the package cosmoprimo 3 3 3 https://github.com/cosmodesi/cosmoprimo . To perform these transforms robustly, we extend the power spectrum models from k_{\mathrm{max}}\sim 0.4 h Mpc -1 to higher values of k using power-law extrapolation. We also apply an exponential damping to the power spectra of the form \exp(-k^{2}a^{2}) , setting a=1\,h^{-1} Mpc, which does not affect any of the parameter fits but mitigates high-frequency oscillations in the outputs. Finally, the geometric factors A(r) and B(r) that appear in Equation 6 and Equation 7 are given by sums over the data sample in each separation bin, A(r)=\frac{\sum_{a