Title: New constraints on MOND from galaxy clusters

Abstract: We revisit the application of Modified Newtonian Dynamics (MOND) to galaxy clusters. We confront the high quality X-ray data for eight clusters of galaxies observed by the \xmm satellite with the predictions of MOND. We obtain a ratio of the MOND dynamical mass to the baryonic mass of $M_m/M_b=4.94\pm 0.50$ in the outer parts (i.e $r\sim 0.5$ ~ R$_{vir}$), in the concordance cosmological model where the predicted asymptotic ratio, if any baryons are present, is $7.7^{+1.4}_{-1.1}$ (at $r\geq 0.3$ ~ R$_{vir}$). We confirm that the MOND paradigm lowers the discrepancy between the binding mass and the baryonic mass in clusters by a factor of $\sim 1.6$ at about half the virial radius. However, at this radius about 80% of the mass is still missing, and as pointed out by \citet{sanders03}, this necessitates a component of dark baryons or neutrinos in the cluster core. Concerning the neutrino hypothesis, application of the new data requires a minimum neutrino mass of $m_\nu>1.74\pm 0.34$ eV to fill this gap. The corresponding 2$\sigma$ lower limit of $m_\nu>1.06$ eV is marginally inconsistent with the current constraints from the cluster number counts, and from the CMB and large scale structure data. MOND must invoke neutrinos to represent the main component that account for the missing mass problem in clusters.