Low-frequency noise in magnetic tunnel junctions

Abstract

The utilization of electron’s spin and charge in electronic nanostructures triggered a new field called spintronics. A big step forward in spintronics is the observation of a pronounced magnetoresistance (MR) effect in magnetic tunnel junction (MTJ), which results in high sensitivity of MTJ sensors. For applications, not only high sensitivity is demanded but also low noise level is required to improve sensor’s detectivity. Typically, intrinsic noise of MTJ exhibits large magnitude at low frequencies. Therefore, investigation of low-frequency noise (LFN) is critical for achieving highly-sensitive MTJ sensors with suppressed noise and enhanced detectivity. Although abundant works have been conducted, there is a lack of an elaborate analytical model of the magnetic low-frequency noise (MLFN) and a quantitative study on the electrical low-frequency noise (ELFN) contributed from both the electrical random telegraph noise (ERTN) and electrical 1/f noise. These constraints are addressed in this thesis. The MLFN was investigated under varying magnetic field and MTJ’s shape anisotropy. The MTJs with different shape anisotropies, achieved by altering their aspect ratios, possessed distinct demagnetizing factors. Large magnetic noise was correlated with the increase of magnetic phase loss of ferromagnetic layers resulted from the enhancement of magnetization fluctuation. Upon increasing the shape anisotropy, a notable suppression of magnetic phase loss in the antiparallel (AP) state was observed while it exhibited a slight decrease in the parallel (P) state, revealing that the increase of shape anisotropy caused a more significant reduction of magnetization fluctuation in the AP state. These phenomena were computationally validated by constructing a macrospin model to describe the thermally-induced magnetization fluctuation. Based on the model, the effect of shape anisotropy can be extended to other types of in-plane uniaxial anisotropies. The enhancement of in-plane uniaxial anisotropy can effectively control and suppress the MLFN. The ELFN contributed from ERTN and electrical 1/f noise was investigated, both of which were caused by the trapping-detrapping (TD) processes among the electron traps induced by the defects in/near the tunnel barrier (TB). The TD processes tended to be more pronounced with increasing bias current and temperature, giving rise to a notable ERTN or electrical 1/f noise. The activation mechanism of TD processes for the electrical 1/f noise was correlated with the electron hopping events among the localized states (LS) induced by defects. Compared to low order LS hopping processes, the higher order hopping processes can probably facilitate more TD events, resulting in a steeper increase of the electrical 1/f noise. To suppress these noises in MTJ, researchers can aim to inhibit the TD processes by fabricating high-quality TB with fewer defects or optimizing operational parameters, such as bias current and working temperature. Continuous research efforts on improving detectivity of MTJ sensors can also be extended to enhance sensing performance of other types of MR sensors. The proliferation of highly-sensitive MR sensors will open up a wide range of applications and offer more magnetic or magnetic-related information to the Internet of Things, enriching and upgrading the context of smart living. (492 words)