Effect of Spin‐Split Band and its Machine Learning Prediction on Tunnel Magnetoresistance across Organic Magnetic Tunnel Junctions

The effect of spin‐split band energy on tunnel magnetoresistance (TMR) of rubrene‐based magnetic tunnel junction devices (x/rubrene/Co, x = La 2 O 3 , LaMnO 3 , La 0.7 Ca 0.3 MnO 3 (LCMO), La 0.7 Sr 0.3 MnO 3 (LSMO)) using nonequilibrium Green's function is investigated. Parallel ( R P ) and antiparallel ( R AP ) resistances have been observed to be affected by spin‐split band energy (Δ) due to modified spin‐dependent scattering at density of states of corresponding spin orientation. The sharp rise of TMR in La 2 O 3 has been attributed to the inhibition of R P by R AP with exchange energy. Spin‐transfer torque (STT) results indicate minimum torque induced spin damping leading to enhanced TMR at higher Δ values. Machine learning‐based models have been prepared to predict the behavior of different parameters ( R P , R AP , TMR (%), STT‐IN, STT‐OUT) versus voltage graphs with varying spin‐split energy. Among various models, Gaussian process regressor and polynomial regression achieve the highest predicting power capability with lease root mean square error. All the parameters associated with LCMO and LaMnO 3 are supposed to be correctly predicted. TMR (%) and STT‐OUT parameters of LSMO device have been correctly predicted as these values are highly correlated with Δ.