Tunnel Junctions

Tunneling is one of the most fundamental problems in physics that also has broad technological applications. Ferromagnet-insulator-ferromagnet tunneling devices (commonly called magnetic tunnel junctions, MTJs) can be used to expand tunneling studies to include the spin of the tunneling electron, which ultimately has applications in advanced memories like Magnetoresistive Random Access Memory (MRAM). The differential tunneling resistance (dV/dI) of an MTJ depends on the density of states of the ferromagnets in the electrodes. As a result, dV/dI is low when the magnetizations of the ferromagnets are parallel, while dV/dI is relatively high in the antiparallel configuration. For applications, this difference in dV/dI of these two states allows defining a logic basis where high resistance is a 1 and low resistance is a 0.

	Fig 1: Over simplified band structure representation of a magnetic tunnel junction. (a) The parallel state has low tunneling resistance. (b) The density of states bottleneck causes a relatively high resistance in the antiparallel state.

Our most recent work on MTJs (in collaboration with Jon Slaughter and Renu Dave from Freescale Semiconductor and Johan Akerman from KTH in Sweden) investigated

• an anomalous bias dependence of the conductance in CoFeB/MgO/CoFeB junctions,
• the impact of interfacial roughness on tunneling conductance and the tunnel barrier parameters extracted using traditional tunneling models, and
• a dynamic resonant tunneling effect in CoFeB/MgO/NiFe junctions.

In the first, we showed that our observation of parabolic tunneling conductance out to nearly 2V can be explained by interfacial roughness. In the second, we showed that interfacial roughness leads to erroneous barrier parameters when data are fit with models that assume the barrier is perfectly smooth. In the third, we discovered a novel effect where electrons tunnel directly into the conduction band of the MgO and undergo interference that results in oscillations of the tunneling conductance. These projects incorporated a significant amount of modeling to complement our interpretations of these data.

Fig 2: The oscillations in the antiparallel dV/dI data (blue) for positive biases is due to electronic interference within the conduction band of the barrier, as illustrated in the inset. The data are from a CoFeB/MgO/NiFe device.

[1] C. W. Miller et al., PRB 74, 212404 (2006)

[2] C. W. Miller et al., APL 90, 043513 (2007)

[3] C. W. Miller et al., PRL (submitted April, 2007)