We propose that the observed small (100 ps) spin relaxation time in graphene is due to resonant scattering by local magnetic moments. At resonances, magnetic moments behave as spin hot spots: the spin-flip scattering rates are as large as the spin-conserving ones, as long as the exchange interaction is greater than the resonance width. Smearing of the resonance peaks by the presence of electron-hole puddles gives quantitative agreement with experiment, for about 1 ppm of local moments. Although magnetic moments can come from a variety of sources, we specifically consider hydrogen adatoms, which are also resonant scatterers. The same mechanism would also work in the presence of a strong local spin-orbit interaction, but this would require heavy adatoms on graphene or a much greater coverage density of light adatoms. To make our mechanism more transparent, we also introduce toy atomic chain models for resonant scattering of electrons in the presence of a local magnetic moment and Rashba spin-orbit interaction. DOI: 10.1103/PhysRevLett.112.116602 PACS numbers: 72.80.Vp, 72.25.Rb Graphene [1,2] has been considered an ideal spintronics [3,4] material. Its spin-orbit coupling being weak, the spin lifetimes of Dirac electrons are expected to be long, on the order of microseconds [5]. Yet experiments find tenths of a nanosecond [6][7][8][9][10][11][12][13]. This vast discrepancy has been the most outstanding puzzle of graphene spintronics. Despite intense theoretical efforts [14][15][16][17][18][19][20][21], the mechanism for the spin relaxation in graphene has remained elusive. Recently, mesoscopic transport experiments [22] found evidence that local magnetic moments could be the culprits. Here we propose a mechanism of how even a small concentration of such moments can drastically reduce the spin lifetime of Dirac electrons. If the local moments sit at resonant scatterers, such as vacancies [23][24][25] and adatoms [25,26], they can act as spin hot spots [27]: while contributing little to momentum relaxation, they can dominate spin relaxation. Our mechanism is general, but to obtain quantitative results we use model parameters corresponding to hydrogen adatoms which yield both resonant scattering and local moments [26,28,29]. The calculated spin relaxation rates for 1 ppm of local moments, when averaged over electron density fluctuations due to electron-hole puddles, are in quantitative agreement with experiment. Our theory shows that in order to increase the spin lifetime in graphene, local magnetic moments at resonant scatterers need to be chemically isolated or otherwise eliminated.In graphene the presence of local magnetic moments is not obvious, unless the magnetic sites (vacancies or adatoms) [30] are intentionally produced [24,25]. It is reasonable to expect that there are not more magnetic sites than, say, 1 ppm, in "clean" graphene samples investigated for spin relaxation in experiments [6][7][8][9][10][11][12][13]. For this concentration a simple estimate gives a weak spin relaxation rate, similar to wha...