Role of boron diffusion in CoFeB/MgO magnetic tunnel junctions

Several scientific issues concerning the latest generation read heads for magnetic storage devices, based on CoFeB/MgO/CoFeB magnetic tunnel junctions (MTJs) are known to be controversial, including such fundamental questions as to the behavior and the role of B in optimizing the physical properties of these devices. Quantitatively establishing the internal structures of several such devices with different annealing conditions using hard x-ray photoelectron spectroscopy, we resolve these controversies and establish that the B diffusion is controlled by the capping Ta layer, though Ta is physically separated from the layer with B by several nanometers. While explaining this unusual phenomenon, we also provide insight into why the tunneling magnetoresistance (TMR) is optimized at an intermediate annealing temperature, relating it to B diffusion, coupled with our studies based on x-ray diffraction and magnetic studies.

Figure 1

Normalized photoelectron spectra of (a) Fe $2p$ and (b) B $1s$ collected from the as-grown trilayer MTJ sample T at photon energies ranging from 3000 to 5000 eV. Panels (c) and (d) show comparisons of Fe $2p$ and B $1s$ spectra obtained at 4000 eV photon energy from the as-grown (black line) and $300{}^{\circ }\mathrm{C}$-annealed (red dots) samples. Photon energy dependencies of normalized photoelectron spectra of Fe $2p$ and B $1s$ collected from the $300{}^{\circ }\mathrm{C}$-annealed sample T are shown in the insets of panels (c) and (d), respectively.

Figure 2

(a) Intensity ratios between different core-level photoelectron spectra obtained at different photon energies from the trilayer sample T, before (circles) and after (diamonds) annealing the sample at $300{}^{\circ }\mathrm{C}$. (b) The average TMR (averaged over the whole wafer) and the resistance area product (RA) obtained from sample T under different annealing conditions. (c) The thickness and nature of various layers in the capped trilayer sample T, as deduced from the analysis of core-level intensities as a function of the photon energy.

Figure 3

(a, b) Fe $2p$ and B $1s$ core-level spectra collected from the uncapped bilayer sample B1 at 3000 eV photon energy before (black) and after (red) annealing the sample at $300\pm 5{}^{\circ }\mathrm{C}$ for 20 min. (c, d) The same spectra for the Ta-capped bilayer sample B2 collected at 3600 eV. The inset of (a) shows the intensity ratio between B oxide to B metal (red triangles) and B oxide to Mg (blue square) obtained from the $300\pm 5{}^{\circ }\mathrm{C}$ annealed sample B1 at different photon energies. The lines represent the intensity ratios for the best fitted structure, shown in (f). The thickness and nature of various layers in the unannealed and annealed uncapped bilayer sample B1 are shown in (e) and (f). The numbers shown in the figures are in nanometers.

Figure 4

Comparison of Fe $2p$ (a) and B $1s$ (b) spectra, collected for trilayer sample T at 4000 eV photon energy before and after annealing the sample in situ at $420{}^{\circ }\mathrm{C}$ for 20 min and subsequently to $520{}^{\circ }\mathrm{C}$ for 20 min. (c, d) The effect of annealing on the x-ray diffraction pattern and on the magnetization, respectively. The inset in panel (d) shows that the perpendicular anisotropy is removed upon annealing at high annealing temperature.