KiDS plus VIKING plus GAMA: Halo occupation distributions and correlations of satellite numbers with a new halo model of the galaxy-matter bispectrum for galaxy-galaxy-galaxy lensing

Context. Halo models and halo occupation distributions (HODs) are important tools to model the distribution of galaxies and matter. Aims. We present and assess a new method for constraining the parameters of HODs using the mean gravitational lensing shear around galaxy pairs, so-called galaxy-galaxy-galaxy lensing (G3L). In contrast to galaxy-galaxy lensing, G3L is also sensitive to the correlations between the per-halo numbers of galaxies from different populations. We employed our G3L halo model to probe these correlations and test the default hypothesis that they are negligible. Methods. We derived a halo model for G3L and validated it with realistic mock data from the Millennium Simulation and a semi-analytic galaxy model. Then, we analysed public data from the Kilo-Degree Survey (KiDS), the VISTA Infrared Kilo-Degree Galaxy Survey (VIKING) and data from the Galaxy And Mass Assembly Survey (GAMA) to infer the HODs of galaxies at z < 0:5 in five different stellar mass bins between 10(8:5)h(-2) M-circle dot and 10(11:5)h(-2) M-circle dot and two colours (red and blue), as well as correlations between satellite numbers. Results. The analysis accurately recovers the true HODs in the simulated data for all galaxy samples within the 68% credibility range. The model best fits agree with the observed G3L signal on the 95% confidence level. The inferred HODs vary significantly with colour and stellar mass. In particular, red galaxies prefer more massive halos greater than or similar to 10(12) M-circle dot, while blue galaxies are present in halos greater than or similar to 10(11) M-circle dot. There is strong evidence (>3 sigma) for a high correlation, increasing with halo mass, between the numbers of red and blue satellites and between galaxies with stellar masses below 10(10) M-circle dot. Conclusions. Our G3L halo model accurately constrains galaxy HODs for lensing surveys of up to 10(3) deg(2) and redshift below 0:5 probed here. Analyses of future surveys may need to include non-Poisson variances of satellite numbers or a revised model for central galaxies. Correlations between satellite numbers are ubiquitous between various galaxy samples and are relevant for halos with masses greater than or similar to 10(13) M-circle dot, that is, of galaxy-group scale and more massive. Possible causes of these correlations are the selection of similar galaxies in different samples, the survey flux limit, or physical mechanisms such as a fixed ratio between the satellite numbers of distinct populations. The decorrelation for halos with smaller masses is probably an effect of shot noise by low-occupancy halos. The inferred HODs can be used to complement galaxy-galaxy lensing or galaxy-clustering HOD studies or as input to cosmological analyses and improved mock galaxy catalogues.