Competition between cotunneling, Kondo effect, and direct tunneling in discontinuous high-anisotropy magnetic tunnel junctions

The transition between Kondo and Coulomb blockade effects in discontinuous double magnetic tunnel junctions is explored as a function of the size of the CoPt magnetic clusters embedded between AlO${}_x$ tunnel barriers. A gradual competition between cotunneling enhancement of the tunneling magnetoresistance (TMR) and the TMR suppression due to the Kondo effect has been found in these junctions, with both effects having been found to coexist even in the same sample. It is possible to tune between these two states with temperature (at a temperature far below the cluster blocking temperature). In addition, when further decreasing the size of the CoPt clusters, another gradual transition between the Kondo effect and direct tunneling between the electrodes takes place. This second transition shows that the spin-flip processes found in junctions with impurities in the barrier are in fact due to the Kondo effect. A simple theoretical model able to account for these experimental results is proposed.

Figure 1

(a) STM images of the CoPt nanoclusters for different nominal layer thicknesses $t$. (b) Average cluster volume estimated by fitting the histogram of the cluster sizes to a log-normal distribution, as shown in the inset.

Figure 2

Blocking temperature $T_B$ (determined from SQUID FC and ZFC measurements) vs nominal layer thickness $t$ of CoPt and CoFe (CoFe data from Ref. 10 by Yang et al.). The inset shows the magnetic moment per atom—from SXMCD—vs $t$. The dotted line is a guide to the eye.

Figure 3

Magnetotransport data: resistance $R$ at a saturating magnetic field of 8 T and TMR of DMTJs against $V$ and $T$ for different nominal CoPt thicknesses $t$. For $t=1.1$ nm there is an enhancement of $R$ (a) and TMR [(b) and c)] due to co-tunneling; for $t=0.7$ nm, $R$ is also enhanced (d) and we see TMR enhancement or suppression depending on $T$ [(e) and (f)]. Offsets have been added for clarity in (e). Resistances at 2.5 K for $t=0$, $0.2$ and $0.3$ nm are shown in (g). For $t=0.3$ nm, (h) a strong Kondo TMR suppression is observed at $T=2.5$ K (dynamic conductance measurements are shown in the inset); for $t=0$ and $0.2$ nm (i) we observe a TMR reduction due to spin scattering at 2.5 K.

Figure 4

Magnetoresistance measurements. (a) TMR of DMTJs vs $V$ for $t=1.1$, $0.7$, and $0.3$ nm at $2.5$ K. When reducing the CoPt thickness, the TMR varies from being enhanced to being suppressed at low $V$. (b) TMR vs $V$ for $t=0.3$, $0.2$, and 0 nm at $2.5$ K. With thicknesses below $t=0.3$ nm, the TMR is recovered. (c) The amplitude of the TMR at low bias, defined as the difference on the TMR when applying 0 and 10 mV, respectively, as a function of $T$ for $t=0.7$ nm. Two different regimes are clearly identified.

Figure 5

Theoretical model. (a) Total conductivity ($\sigma ={\sigma }_K+{\sigma }_E$), elastic conductivity through the clusters without spin flips (${\sigma }_E$), and conductivity through the clusters showing the Kondo effect (${\sigma }_K$) vs temperature. (b) Relative contribution to $\sigma$ of ${\sigma }_E$ and ${\sigma }_K$ as a function of temperature. (c) Evolution of the TMR with temperature considering and not considering the interplay of the Kondo effect: in the first case ${\sigma }_K$ follows Eq. (2), while in the second one ${\sigma }_K=0$.