We propose that the observed spin relaxation in bilayer graphene is due to resonant scattering by magnetic impurities. We analyze a resonant scattering model due to adatoms on both dimer and nondimer sites, finding that only the former give narrow resonances at the charge neutrality point. Opposite to singlelayer graphene, the measured spin-relaxation rate in the graphene bilayer increases with carrier density. Although it has been commonly argued that a different mechanism must be at play for the two structures, our model explains this behavior rather naturally in terms of different broadening scales for the same underlying resonant processes. Not only do our results-using robust and first-principles inspired parameters-agree with experiment, they also predict an experimentally testable sharp decrease of the spin-relaxation rate at high carrier densities. DOI: 10.1103/PhysRevLett.115.196601 PACS numbers: 72.80.Vp, 72.25.Rb Understanding spin relaxation is essential for designing spintronics devices [1,2]. Unfortunately, spin relaxation in graphene structures has been a baffling problem [3]. While experiments in both single layer graphene (SLG) [4][5][6][7][8][9][10][11] and bilayer graphene (BLG) [7,8] yield spin lifetimes on the 100-1000 ps time scale (the highest values achieved in graphene/h-BN structures [12,13]), theories based on realistic spin-orbit coupling and transport parameters predict lifetimes on the order of microseconds [14][15][16][17][18][19][20][21][22][23][24].While the magnitudes of the spin-relaxation rates of SLG and BLG are similar, the dependence of the rates on the electron density is opposite in the two systems. In SLG the spin-relaxation rate decreases with increasing the carrier density [5][6][7][8], in BLG the spin-relaxation rate increases [7,8]. Since the diffusivity in the investigated samples decreases with increasing the electron density, it has been a common practice to assign two different mechanisms to both structures: the Elliott-Yafet mechanism [25,26] to SLG [5,6,9,10] and Dyakonov-Perel mechanism [27] to BLG [7][8][9].The main problem with that assignment is quantitative. Spin-orbit coupling in graphene [28] is too weak to yield such a small spin-relaxation time. An explicit first-principles calculation [23] predicts that one would need 0.1% of adatoms to give a 100 ps spin lifetime. Recently a new mechanism for SLG was proposed [29] (see also Ref.[30]), based on resonant scattering off local magnetic moments. It gives the observed spin-relaxation times with as little as 1 ppm of local magnetic moments and also agrees with the experimental behavior for SLG of decreasing the spinrelaxation rate with increasing electron density. Where do these local moments come from? It was theoretically predicted that adatoms such as hydrogen [31,32], but also chemisorbed organic molecules [33] can be responsible. Experimentally it was demonstrated that hydrogen adatoms indeed induce local moments [34,35], but even untreated graphene flakes were shown to exhibit 20 ppm spin 1=2 para...