## Abstract

The process of grain growth has been investigated for four close-packed hexagonal metals magnesium, zinc, cadmium and thallium. The tests were carried out with thin sheets of metal, at least 10 grains in thickness, which were cross-rolled and allowed to recrystallize completely before being raised to the temperature at which the measurements of grain diameter were made, the surface being etched before each grain count. It has been established that maintenance at a fixed temperature leads ultimately to a stable average grain diameter D$_e$ characteristic of the temperature, supposing the initial value of diameter to have been less than D$_e$. The relation between D$_e$ and temperature T is D$^\frac{1}{2}_e$ = C(T-T$_0$), where C and T$_0$ are characteristic of the metal. The way in which grain size increases with time falls into three stages. If the initial grain diameter D$_i$ is very small compared to D$_e$, very rapid growth takes place: in an extreme case at high temperature a threefold increase of diameter in 1 s was established. If D$_i$ is not smaller than 0.65 D$_e$ the observed growth takes place, in general over a period of days, in close accordance with the law D = D$_e$(1 + pe$^{-qt}$)$^{-1}$. This we call the normal stage of growth, to which most attention has been devoted. There is an intermediate stage which characterizes most of the measurements hitherto made by others. For the normal stage q = ae$^{-\epsilon/kT}$, giving an activation energy $\epsilon$ which is 2.25 times the latent heat of fusion. The statistical distribution of grain size is given by z/z$_m$ = $\exp$[-$\pi^2${(x/x$_m$)$^{\frac{1}{2}}$-1}$^2$], where z is the number of grains of diameter x, z$_m$ being the maximum value of z at x$_m$. If the temperature is raised, the number of grains decreasing in diameter is 5.25 times the number increasing in diameter; 5.25 is about the average number of corners of a grain as seen in the plane section. Measurements have been made on the effect of external stress on grain growth. In the discussion of the results the D$^\frac{1}{2}_e$ law is attributed to the existence of strips of good fit on the boundary between random crystal grains, and an attempt is made to account for grain growth in terms of the disappearance of small grains as a result of surface tension forces and of the rotation of the boundaries of large grains to a position of stability. It is suggested that if a metal could be brought to an amorphous state at T$_0$ it would be stable in that state at that temperature, in which connexion quenching condensation receives reference.