The distribution of the easy-axes in an array of MRAM cells is a vital parameter to understand the switching and characteristics of the devices. By measuring the coercivity as a function of applied-field angle, and remaining close to the perpendicular orientation, a classic Stoner-Wohlfarth approximation has been applied to the resulting variation to determine the standard deviation, σ\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma $$\end{document}, of a Gaussian distribution of the orientation of the easy-magnetisation directions. In this work we have compared MRAM arrays with nominal cells sizes of 20 nm and 60 nm and a range of free layer thicknesses. We have found that a smaller diameter cell will have a wider switching-field distribution with a standard deviation σ=9.5∘\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma = {9.5}^{\circ }$$\end{document}. The MRAM arrays consist of pillars produced by etching a multilayer thin film. This value of σ\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma $$\end{document} is dominated by pillar uniformity and edge effects controlling the reversal, reinforcing the need for ever-improving etch processes. This is compared to larger pillars, with distributions as low as σ=5.5∘\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma = {5.5}^{\circ }$$\end{document}. Furthermore we found that the distribution broadens from σ=5.5∘\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma = {5.5}^{\circ }$$\end{document} to σ=8.5∘\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$\sigma = {8.5}^{\circ }$$\end{document} with free layer thickness in larger pillars and that thinner films had a more uniform easy-axis orientation. For the 20 nm pillars the non-uniform size distribution of the pillars, with a large and unknown error in the free-layer volume, was highlighted as it was found that the activation volume for the reversal of the free layer 930 nm3\documentclass[12pt]{minimal}
\usepackage{amsmath}
\usepackage{wasysym}
\usepackage{amsfonts}
\usepackage{amssymb}
\usepackage{amsbsy}
\usepackage{mathrsfs}
\usepackage{upgreek}
\setlength{\oddsidemargin}{-69pt}
\begin{document}$$^3$$\end{document} was larger than the nominal physical volume of the free layer. However for the 60 nm pillars, the activation volume was measured to be equal to one fifth of their physical volume. This implies that the smaller pillars effectively reverse as one entity while the larger pillars reverse via an incoherent mechanism of nucleation and propagation.
