Hard X-ray Standing-wave Photoelectron Spectroscopy Study of CoFeB/MgO Magnetic Tunnel Junction Multilayers

Book excerpt: As one key aspect of the area of spin-based electronics or spintronics, the magnetic tunnel junction (MTJ) holds special promise for magnetic memory, and possibly also logic devices. In an MTJ, two ferromagnetic layers are separated by a very thin nonmagnetic insulating layer and the key effect is based on the spin-dependent tunneling of electrons through the insulating layer and is called tunnel magnetoresistance (TMR). Resistance is lower when the two ferromagnetic layers are oriented parallel to one another, and higher when they are anti-parallel. Recent work reveals that MTJs with a Ta/CoFeB/MgO/CoFeB/Ta structure show three optimal characteristics: 1) high thermal stability on the nanoscale, 2) a high TMR ratio, and 3) low switching current for current-induced switching of magnetization across the interface. Studies suggest that B diffusion from the initially amorphous CoFeB layer into the MgO causes CoFeB crystallization such that TMR-increasing perpendicular anisotropy (PMA) arises at the MgO/CoFeB interface. Furthermore, the TMR ratio is likewise regulated by B diffusion into the Ta layer. Scientists are currently exploring the structure/properties relationships of these buried interfaces in magnetic nanostructures. One effective method for probing the composition, structure, and properties of buried layers is the newly-developed technique of standing-wave, hard x-ray photoelectron spectroscopy or SW-HAXPES. In this method, a standing wave is generated by Bragg reflection from a synthetic multilayer mirror upon which the sample is deposited. This standing wave can be scanned vertically through the sample by varying the incidence angle around the Bragg angle, giving a rocking curve (RC) scan.Using SW-HAXPES, we studied the B distribution in a Ta/Co0.2Fe0.6B0.2/MgO sample. We obtained hard x-ray standing-wave data, as well as conventional variable takeoff angle XPS (angle-resolved XPS or ARXPS) data at SPring-8 in Japan, as well as complimentary soft x-ray ARXPS data at the Lawrence Berkeley National Laboratory. We compared experimental XPS and rocking-curve (RC) results with results from theoretical calculations. In addition, we obtained ALS (Advanced Light Source at Lawrence Berkeley National Laboratory) reflectance measurements showing us the Bragg angle and helping us to understand the internal structure of the sample. In our results, we found that significant B remains in the CoFeB layer after an optimal annealing treatment, and that outward diffusion of B occurs very nearly uniformly into a thin MgO layer of 10-20 ©5 thickness. Our comparison of theoretical with experimental RCs revealed that the boron diffused from both the top and bottom sides of the CoFeB interface. Corroborating this finding, our XPS data revealed chemical-shifted B 1s and Ta 4f peaks. Measurement of the ratio of chemical-shift peak intensity over non-chemical shift peak intensity for B 1s and Ta 4f B1s indicated the amount of B that diffused into the MgO and Ta layers respectively, which assisted us in arriving at our model for the sample. We were careful to conserve the total number of B atoms in our calculations for the final best-fit sample geometry. Our findings indicated uniform boron distribution through the postannealed MgO layer, rather than a boron concentration gradient. Confirming these findings, the rocking-curve 3D intensity maps showed a chemical shift for B 1s, O 1s, and Mg 1s peaks, although we did not see a chemical shift for the Ta 4p peak due to limitation in the range of our data.Thus, we have at our disposal a method for studying the structure/properties relationship in MTJs and related spintronic devices which can be used in conjunction with other methods for the development of these materials for TMR and low switching-current applications. This dissertation has thus demonstrated a new tool for the optimization of future spintronic materials.