Euclid preparation. BAO analysis of photometric galaxy clustering in configuration space

With about 1.5 billion galaxies expected to be observed, the very large number of objects in the Euclid photometric survey will allow for precise studies of galaxy clustering from a single survey, over a large range of redshifts $0.2 < z < 2.5$. In this work, we use photometric redshifts to extract the baryon acoustic oscillation signal (BAO) from the Flagship galaxy mock catalogue with a tomographic approach to constrain the evolution of the Universe and infer its cosmological parameters. We measure the two-point angular correlation function in 13 redshift bins. A template-fitting approach is applied to the measurement to extract the shift of the BAO peak through the transverse Alcock--Paczynski parameter $\alpha$. A joint analysis of all redshift bins is performed to constrain $\alpha$ at the effective redshift $z_\mathrm{eff}=0.77$ with MCMC and profile likelihood techniques. We also extract one $\alpha_i$ parameter per redshift bin to quantify its evolution as a function of time. From these 13 $\alpha_i$, which are directly proportional to the ratio $D_\mathrm{A}/\,r_\mathrm{s,\,drag}$, we constrain $h$, $\Omega_\mathrm{b}$, and $\Omega_\mathrm{cdm}$. From the joint analysis, we constrain $\alpha(z_\mathrm{eff}=0.77)=1.0011^{+0.0078}_{-0.0079}$, which represents a three-fold improvement over current constraints from the Dark Energy Survey. As expected, the constraining power in the analysis of each redshift bin is lower, with an uncertainty ranging from $\pm\,0.13$ to $\pm\,0.024$. From these results, we constrain $h$ at 0.45 %, $\Omega_\mathrm{b}$ at 0.91 %, and $\Omega_\mathrm{cdm}$ at 7.7 %. We quantify the influence of analysis choices like the template, scale cuts, redshift bins, and systematic effects like redshift-space distortions over our constraints both at the level of the extracted $\alpha_i$ parameters and at the level of cosmological inference.