Intrinsic correlations of galaxy sizes in a hydrodynamical cosmological simulation

Residuals between measured galactic radii and those predicted by the Fundamental Plane (FP) are possible tracers of weak lensing magnification. However, observations have shown these to be systematically correlated with the large-scale structure. We use the Horizon-AGN hydrodynamical cosmological simulation to analyse these intrinsic size correlations (ISCs) for both elliptical (early-type) and spiral (late-type) galaxies at $z=0.06$. We fit separate FPs to each sample, finding similarly distributed radius residuals, $\lambda$, in each case. We find persistent $\lambda\lambda$ correlations over three-dimensional separations $0.5-17\,h^{-1}{\rm{Mpc}}$ in the case of spiral galaxies, at $>3\sigma$ significance. When relaxing a mass-selection, applied for better agreement with galaxy clustering constraints, the spiral $\lambda\lambda$ detection strengthens to $9\sigma$; we detect a $5\sigma$ density-$\lambda$ correlation; and we observe intrinsically-large spirals to cluster more strongly than small spirals over scales $\lesssim10\,h^{-1}{\rm{Mpc}}$, at $>5\sigma$ significance. Conversely, and in agreement with the literature, we observe lower-mass, intrinsically-small ellipticals to cluster more strongly than their large counterparts over scales $0.5-17\,h^{-1}{\rm{Mpc}}$, at $>5\sigma$ significance. We model $\lambda\lambda$ correlations using a phenomenological non-linear size model, and predict the level of contamination for cosmic convergence analyses. We find the systematic contribution to be of similar order to, or dominant over the cosmological signal. We make a mock measurement of an intrinsic, systematic contribution to the projected surface mass density $\Sigma(r)$ and find statistically significant, low-amplitude, positive (negative) contributions from lower-mass spirals (ellipticals), which may be of concern for large-scale ($\gtrsim\,7\,h^{-1}$ Mpc) measurements.