In tunnel junctions between ferromagnets and heavy elements with strong spin orbit coupling the magnetoresistance is often dominated by tunneling anisotropic magnetoresistance (TAMR). This makes conventional DC spin injection techniques impractical for determining the spin relaxation time (τs). Here, we show that this obstacle for measurements of τs can be overcome by 2nd harmonic spin-injection-magnetoresistance (SIMR). In the 2nd harmonic signal the SIMR is comparable in magnitude to TAMR, thus enabling Hanle-induced SIMR as a powerful tool to directly determine τs. Using this approach we determined the spin relaxation time of Pt and Ta and their temperature dependences. The spin relaxation in Pt seems to be governed by Elliott-Yafet mechanism due to a constant resistivity×spin relaxation time product over a wide temperature range.PACS numbers: 72.25. Rb, 72.25.Ba, 73.50.Bk, 73.40.Rw The large applied potential of spin-orbit-torques for magnetic random access memory has stimulated intensive interest in investigating spin orbit coupling (SOC) in heavy metals such as Pt and Ta [1][2][3][4][5][6][7][8][9][10][11]. Their spin Hall angle (θ SH ), spin diffusion length (l s ) and spin relaxation time (τ s ), which influence switching efficiency are important parameters for determining their effectiveness, but especially the latter two are experimentally hard to assess. Accurate determination of τ s could also help to identify the spin relaxation mechanisms [12]. Though θ SH and l s have been measured by spin pumping [13][14][15][16][17] In fact, spin injection experiments in nonlocal spin valves [29][30][31][32][33][34][35] and 3-terminal geometries [36][37][38][39][40] are both powerful tools in measuring τ s in metals and semiconductors. In these experiments, ferromagnetic layer (FM)/tunnel barrier/nonmagnetic layer (NM) junctions are adopted to both inject a non-equilibrium spin accumulation and simultaneously determine their magnitude. These measurement were used to determine spin relaxation times in a wide variety of materials, e.g., τ s,Si =55 -285 ps for heavily doped silicon [40]; τ s,Graphene > 1 ns for graphene/BN [41]; τ s,Al =110 ps for aluminum [29], τ s,Cu =22 ps for copper [42] and τ s,Au =45 ps for gold [32].However, it is impractical to apply these spin injection experiments to measure τ s in heavy metals with strong SOC for at least two reasons. First, l s in this case is so short (about several nanometers) that the preparation of nonlocal spin valves with comparable dimensions is beyond current lithography capabilities. Second, the real contact resistance is r = r C + r SI , where r SI and r C are the contact resistance induced by spin injection (SI) and the original contact resistance without r SI , respectively. Here r SI equals to r N r C /(r N +r C ) and the spin resistance in the NM layer r N is defined as ρ N l sN . ρ N and l sN are the resistivity and spin relaxation length of NM, respectively. For this discussion we ignore the influence of spin resistance in FM on r SI due to...