Complexities in striations of Czochralski silicon crystals are explored on the basis of timehyphen;series analysis by neural networks and by nonlinear regression of a simplex projection method. Two categories of the time series of striations along the crystal growth axis, observed using xhyphen;ray topograph, are forecasted by the simplex projection method and characterized in terms of the dependence of prediction accuracy on embedding dimension and predictionhyphen;time interval. The time series have shorthyphen;term predictability, independent of embedding dimension. The backhyphen;propagation network that has learned the dynamics in one time series can make predictions about the other striation. These results suggest that there exist underlying nonlinear dynamics common to both the striations.
展开▼
